CN110287029A - A method of it is adjusted based on kubernetes container resource dynamic - Google Patents
A method of it is adjusted based on kubernetes container resource dynamic Download PDFInfo
- Publication number
- CN110287029A CN110287029A CN201910566915.1A CN201910566915A CN110287029A CN 110287029 A CN110287029 A CN 110287029A CN 201910566915 A CN201910566915 A CN 201910566915A CN 110287029 A CN110287029 A CN 110287029A
- Authority
- CN
- China
- Prior art keywords
- usage amount
- pod
- resource
- resource usage
- container
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
- G06F2009/45583—Memory management, e.g. access or allocation
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Debugging And Monitoring (AREA)
Abstract
The invention discloses a kind of methods based on the resource dynamic adjustment of kubernetes container, comprising: passes through the real-time container real resource usage amount of the index collection of server built in kubernetes to pod;It according to the real resource usage amount being collected into, is integrated, calculates separately and record the target resource usage amount of the container of the recommendation of pod;The resource situation information for persistently monitoring pod, pod is restarted;Request to create is intercepted when pod is restarted, and preceding aim resource usage amount is updated into the resource distribution of pod, and new pod is created with the configuration.Method based on the resource dynamic adjustment of kubernetes container of the invention, the resource request usage amount of kubernetes cluster minimum load unit pod is adjusted by dynamic, it can reduce the risk that container uses up memory and cpu starvation, improve the resource utilization of entire kubernetes cluster.
Description
Technical field
The present invention relates to the operation control technology fields of software application, especially a kind of to be based on kubernetes container resource
The method of dynamic adjustment.
Background technique
With the continuous development of cloud computing, container and Kubernetes have become the foundation stone of cloud native applications,
Kubernetes is becoming vast Internet company and tradition IT industry carries out cloudization and simplifies the sharp weapon of O&M, in production ring
Large scale deployment is obtained on border and is used by more and more companies.
Application load resource distribution is manually in pod creation just in advance into configuring one on existing kubernetes
As will not all change in the life cycle entirely applied, such as need to change, restart pod after manually carrying out configuration modification, deposit
In the disadvantage too strong to people's dependence.Also there is the horizontal automatic telescopic device (HPA) of the resource management pod based on software performance index,
According to the cpu of currently running pod occupy and EMS memory occupation come the quantity of dynamic retractility pod, exist be not adapted dynamically it is each
The shortcomings that resource occupation amount of pod, the quantity for only adjusting pod can not be accurately controlled the hardware resource utilization rate of node.
Summary of the invention
Goal of the invention of the invention is, in view of the above-mentioned problems, providing a kind of based on kubernetes container resource dynamic tune
Whole method is the container dispatching method with time correlation based on Kubernetes container cluster management platform.
In order to achieve the above objectives, the technical scheme adopted by the invention is that:
A method of it is adjusted based on kubernetes container resource dynamic, the following steps are included:
S1, every preheating setting time, pass through the real-time container of the index collection of server built in kubernetes to pod
Real resource usage amount;
S2, according to the real resource usage amount being collected into step S1, integrated, calculate separately and record pushing away for pod
The target resource usage amount for the container recommended;
S3, the resource situation information for persistently monitoring pod, when the target resource usage amount of pod is configured higher or lower than container
When the presetting numerical value of resource usage amount, or when memory occur within the presetting period use up the event for being forced to delete
When, pod is restarted;
S4, request to create is intercepted when pod is restarted, and preceding aim resource usage amount is updated into the resource distribution to pod
In, and new pod is created with the configuration.
As an option, step S2 specifically includes the following contents:
S21, the service condition information including CPU and memory usage amount for obtaining each pod in cluster, and carry out whole
It closes;
S22, using margin algorithm and confidence algorithm, respectively obtain target resource usage amount, the usage amount upper limit
And usage amount lower limit;It is specific as follows:
Target resource usage amount is calculated using margin algorithm, and margin algorithmic formula is as follows:
TargetRes ource=RecourceAmount* (1+marginFraction), wherein
RecourceAmount is the real resource usage amount being collected into, and marginFraction is target resource resource usage amount than real
The multiple that border resource usage amount has more;
The usage amount upper limit and usage amount lower limit all substitute into the superposition of confidence algorithm using by the result of margin algorithm
It calculates and obtains;Confidence algorithmic formula is as follows:
ScaledResource=originalResource* (1+multiplier/confidence) ^exponent;
Wherein, originalResource is the result TargetResource of aforementioned margin algorithm;confidence
For the data duration of acquisition;The usage amount upper limit and the multiplier value of usage amount lower limit are respectively 1 and 0.001;In usage amount
Limit and the exponent value of usage amount lower limit are respectively 1 and -2;
S23, the target resource usage amount, the usage amount upper limit and usage amount lower numerical limit of the container of the recommendation of pod are updated
It is stored in status field, prepares for down-stream.
As an option, step S3 specifically includes the following contents:
S31, the resource situation information that a pod is obtained every the t1 time;
S32, the event information that pod is obtained by the status field information of pod, as the presetting t2 started in pod
Occurred to then follow the steps S35 when memory uses up the event for being forced to delete in period, it is no to then follow the steps S33;
S33, configuration resource usage amount information is obtained by the resource field information of pod, when the configuration resource of pod makes
S35 is thened follow the steps when dosage is not between the aforementioned usage amount upper limit and usage amount lower limit, it is no to then follow the steps S34;
S34, the configuration resource usage amount of pod and preceding aim resource usage amount are compared, when configuration resource uses
Measuring then terminates process when being less than presetting numerical value k with the gap ratio of target resource usage amount and exits, no to then follow the steps
S35;
S35, pod is carried out to chase operation, executes pod reboot operation.
As an option, the particular content of step S4 is as follows:
The new pod that creates that S41, interception are sent to kubernetes application programming interface server is requested;
S42, using target resource usage amount as pod resou rce request usage amount update to pod resource distribution
In, and modified pod resource allocation information is sent to kubernetes application programming interface server, it completes new
The creation of pod.
Due to the adoption of the above technical scheme, the invention has the following advantages:
Method based on the resource dynamic adjustment of kubernetes container of the invention, adjusts kubernetes collection by dynamic
The resource request usage amount of group's minimum load unit pod, reduces the cost of manual maintenance;By the inventive method, can reduce
Container uses up the risk of memory and cpu starvation, improves the resource utilization of entire kubernetes cluster.
Detailed description of the invention
Fig. 1 is step block diagram of the invention.
Fig. 2 is the process flow diagram of one embodiment of the invention.
Fig. 3 is the use block diagram of one embodiment of the invention.
Fig. 4 is the structural block diagram of one embodiment of the invention.
Specific embodiment
It is further illustrated below in conjunction with specific implementation of the attached drawing to invention.
As shown in Figure 1, a kind of method based on the resource dynamic adjustment of kubernetes container, comprising the following steps:
Step S1, every preheating setting time, pass through the index collection of server built in kubernetes to the real-time of pod
Container real resource usage amount.Preheating setting time t0 is configurable value, and default value is 1 minute.
Step S2, it according to the real resource usage amount being collected into step S1, is integrated, calculates separately and record pod
Recommendation container target resource usage amount.
In one example, step S2 specifically includes the following contents:
S21, the service condition information including CPU and memory usage amount for obtaining each pod in cluster, and carry out whole
It closes;
S22, using margin algorithm and confidence algorithm, respectively obtain target resource usage amount, the usage amount upper limit
And usage amount lower limit;It is specific as follows:
Target resource usage amount is calculated using margin algorithm, and margin algorithmic formula is as follows:
TargetRes ource=RecourceAmount* (1+marginFraction), wherein
RecourceAmount is the real resource usage amount being collected into, and marginFraction is target resource resource usage amount than real
The multiple that border resource usage amount has more;
The usage amount upper limit and usage amount lower limit all substitute into the superposition of confidence algorithm using by the result of margin algorithm
It calculates and obtains;Confidence algorithmic formula is as follows:
ScaledResource=originalResource* (1+multiplier/confidence) ^exponent;
Wherein, originalResource is the result TargetResource of aforementioned margin algorithm;confidence
For the data duration of acquisition;The usage amount upper limit and the multiplier value of usage amount lower limit are respectively 1 and 0.001;In usage amount
Limit and the exponent value of usage amount lower limit are respectively 1 and -2;
S23, the target resource usage amount, the usage amount upper limit and usage amount lower numerical limit of the container of the recommendation of pod are updated
It is stored in status field, prepares for down-stream.
Step S3, the resource situation information for persistently monitoring pod, when the target resource usage amount of pod is higher or lower than container
When configuring the presetting numerical value of resource usage amount, or when memory occur within the presetting period using up to be forced to delete
When event, pod is restarted.
Step S3 specifically includes the following contents:
S31, the resource situation information that a pod is obtained every the t1 time;
S32, the event information that pod is obtained by the status field information of pod, as the presetting t2 started in pod
Occurred to then follow the steps S35 when memory uses up the event for being forced to delete in period, it is no to then follow the steps S33;
S33, configuration resource usage amount information is obtained by the resource field information of pod, when the configuration resource of pod makes
S35 is thened follow the steps when dosage is not between the aforementioned usage amount upper limit and usage amount lower limit, it is no to then follow the steps S34;
S34, the configuration resource usage amount of pod and preceding aim resource usage amount are compared, when configuration resource uses
Measuring then terminates process when being less than presetting numerical value k with the gap ratio of target resource usage amount and exits, no to then follow the steps
S35;
S35, pod is carried out to chase operation, executes pod reboot operation.
It wherein, is comprising endpoint value between the usage amount upper limit and usage amount lower limit of step S33, e.g., target resource makes
The 90% of dosage is less than configuration resource usage amount and is less than the 110% of target resource usage amount.Gap ratio is target resource use
Amount is higher or lower than the ratio of container configuration resource usage amount.In one example, t1 is configurable value, and default value is 1 minute;
T2 is configurable value, and default value is 10 minutes;K is configurable value, default value 10%.
Step S4, request to create is intercepted when pod is restarted, and preceding aim resource usage amount is updated into the resource to pod
In configuration, and new pod is created with the configuration.
In one example, the particular content of step S4 is as follows:
The new pod that creates that S41, interception are sent to kubernetes application programming interface server is requested;
S42, using target resource usage amount as pod resource request usage amount update to pod resource distribution
In, and modified pod resource allocation information is sent to kubernetes application programming interface server, it completes new
The creation of pod.Wherein, resource be have Resource field at container, then next stage has CPU, MEMORY, then
Next stage is request, that is, the request usage amount of cpu and memory.
Based on the method for aforesaid receptacle resource dynamic adjustment, a kind of vertical automatic telescopic device (VPA) of Pod container group is constituted,
Including acquisition module, nominator's module, renovator module and admission controller module.It is following will be to the vertical automatic telescopic of container group
Explanation is unfolded in the modules of device, and not most place refers to aforementioned.
Acquisition module: for passing through the index service built in kubernetes every preheating setting time (as every 1 minute)
Device collects the real-time container real resource usage amount of pod.
Nominator's module: for being integrated, calculating separately and record pod according to the real resource usage amount being collected into
Recommendation container target resource usage amount.
Renovator module: for persistently monitoring the resource situation information of pod, be higher than when the target resource usage amount of pod or
When configuring the presetting numerical value of resource usage amount (such as gap ratio is less than 10%) lower than container, or when in the presetting period
It is interior occurred (in such as 10 minutes) memory use up be forced delete event when, pod is restarted.
Admission controller module: it is updated for intercepting request to create when pod is restarted, and by preceding aim resource usage amount
Into the resource distribution of pod, and new pod is created with the configuration.
It is following to will be further illustrated.
As shown in figs 2-4.For present embodiment using 3 Master Node as control node, control node is not responsible
Workload is run, the component of only some kubernetes is operated in above in the form of container, including application programming interfaces
Server (API Server), controller management control centre (Controller Manager), scheduler (Scheduler).
API Server on each Master Node can be connect with distributed data base etcd, for each in cluster
Kind resource distribution and state storage.
As shown in figure 3, in dotted line frame be innovation realization of the present invention the vertical automatic telescopic device (VPA) of Pod container group, can
With in cluster arbitrary node run.
Wherein, API Server and Metrics Server is the subsidiary component of kubernetes cluster, can be by cluster
Interior resource service condition is exposed by kubernetes API form, other modules is facilitated to acquire.Pod is
The minimum load unit of kubernetes cluster, the usually application load in cluster.
As shown in Fig. 2, should be comprised the steps of based on the method for kubernetes container resource dynamic adjustment:
Step 1 is executed, can execute this primary step per presetting time interval, recommender (nominator) is from API
Server collects the resource usage amount of pod, passes through the available target of many algorithms (target) usage amount, the usage amount upper limit
(upperbound), usage amount lower limit (lowerbound), and be arranged into VPA API object.It is wherein " presetting
Time interval " period configuring when being recommender starting, be defaulted as 1 minute.
In conjunction with Fig. 4, the step specifically:
Each VPA that the vertical automatic telescopic controller of step 11. container group (VPA Controller) obtains in cluster refers to
To pod resource usage amount situation, including CPU usage amount and memory usage amount, and integrated.VPA resource is provided with pod
Source binding, in general 1 VPA can bind the pod of several same types.
Step 12.Recommender calculates the resource usage amount data after integration by many algorithms, wherein mesh
It marks usage amount and margin algorithm is used only, formula is as follows:
TargetResource=RecourceAmount* (1+marginFraction)
Wherein, marginFraction defaults value 0.15, and RecourceAmount is collected reality in step 11
Resource usage amount.It can be understood as 1.15 times that target resource usage amount is real resource usage amount.
The amount of the being recommended to use upper limit and the calculation method of the amount of being recommended to use lower limit are that the result of margin algorithm substitutes into
Confidence algorithm, wherein the formula of confidence algorithm is as follows:
ScaledResource=originalResource* (1+multiplier/confidence) ^exponent
Wherein, it is one that originalResource, which is the result TargetResource, confidence of margin algorithm,
A value for describing confidence level, for the data duration (day) of acquisition;The multiplier of the usage amount upper limit is 1, usage amount lower limit
Multiplier is 0.001;The exponent of the usage amount upper limit is 1, and the exponent of usage amount lower limit is -2.It is known that adopting
The data duration of collection is longer, and data are more credible, and the bound of resource usage amount is narrower.
It is the use of other modules in the status field that 3 of calculating are recommended numerical value to update deposit VPA by step 13.
It prepares.
Step 2 is executed, renovator (updater) monitors the information of (watch) all pod for possessing VPA, including configuration
Resource request number information, memory exhaust (OOM) event information etc., these comprehensive situations finally judge whether to execute the pod's
Update operation.
In conjunction with Fig. 4, the step specifically:
Renovator inside the vertical automatic telescopic controller of step 21. container group is from the status acquisition of information of the pod pod
The event that memory exhausts (OOM) and is terminated whether occurred in 10 minutes of starting, in case of mistake, executes step 24, such as
Fruit did not occurred, and executed step 22.
If the usage amount of the resource usage amount information of the resource field configuration of step 22.pod not in step 12
Between the upper limit and usage amount lower limit, then step 24 is directly executed, it is no to then follow the steps 23.
The target usage amount recorded in the resource usage amount and step 12VPA of the resource field configuration of step 23.pod
It compares, it is no to then follow the steps 24 if do not operated to pod with target usage amount gap ratio less than 10%.
Pod is expelled (evict) to operate by step 24., and controller of pod such as copy set replicaSet etc. can incite somebody to action
Pod is restarted, so that new resource usage amount comes into force.
As shown in figure 4, vertical automatic telescopic controller admission controller (the VPA Admission of container group
It Controller is) that the one of of kubernetes admission controller MutatingAdmissionWebhook realizes, it
Essence is a network hook (webhook), can first call the webhook to execute one before specified request reaches API Server
A little operations may carry out some changes to the resource distribution of operation, in this embodiment it is that the request to create of pod has been intercepted,
It is added to new resource usage amount information, then carries out the creation of pod.
Execute step 3, vertical automatic telescopic controller admission controller (the VPA Admission of container group
Controller request to create) can be intercepted when pod is restarted, and more by the recommendation resource usage amount being recorded in VPA in step 1
Newly into the resource distribution of pod, and new pod is created with the configuration.
As above-mentioned, the resource request usage amount of kubernetes cluster minimum load unit pod is adjusted by dynamic, it can
The risk that container uses up memory and cpu starvation is reduced, the resource utilization of entire kubernetes cluster is improved.Have following
Advantage:
1, it because Pod resource request value is adapted dynamically, is entirely capable of needed for it, so kubernetes clustered node makes
It is improved with efficiency.
2, Pod can be arranged on the kubernetes clustered node with appropriate available resources.
3, time-consuming benchmark test task need not be run to determine the right value of CPU and memory request.
4, VPA can adjust at any time CPU and memory request according to using operating condition, without manually performing any operation,
So maintenance time is reduced.
Above description is the detailed description for the present invention preferably possible embodiments, but embodiment is not limited to this hair
Bright patent claim, it is all the present invention suggested by technical spirit under completed same changes or modifications change, should all belong to
In the covered the scope of the patents of the present invention.
Claims (10)
1. a kind of method based on the resource dynamic adjustment of kubernetes container, which comprises the following steps:
S1, every preheating setting time, the real-time container for passing through the index collection of server built in kubernetes to pod is practical
Resource usage amount;
S2, according to the real resource usage amount being collected into step S1, integrated, calculate separately and record the recommendation of pod
The target resource usage amount of container;
S3, the resource situation information for persistently monitoring pod configure resource when the target resource usage amount of pod is higher or lower than container
When the presetting numerical value of usage amount, or when memory occurring within the presetting period using up the event for being forced to delete,
Pod is restarted;
S4, request to create is intercepted when pod is restarted, and preceding aim resource usage amount is updated into the resource distribution of pod, and
New pod is created with the configuration.
2. a kind of method based on the resource dynamic adjustment of kubernetes container according to claim 1, it is characterised in that:
The step S2 specifically includes the following contents:
S21, the service condition information including CPU and memory usage amount for obtaining each pod in cluster, and are integrated;
S22, using margin algorithm and confidence algorithm, respectively obtain target resource usage amount, the usage amount upper limit and make
Dosage lower limit;It is specific as follows:
Target resource usage amount is calculated using margin algorithm, and formula is as follows:
TargetResource=RecourceAmount* (1+marginFraction), wherein RecourceAmount is to receive
The real resource usage amount collected, marginFraction are what target resource resource usage amount had more than real resource usage amount
Multiple;
The usage amount upper limit and usage amount lower limit all substitute into confidence algorithm superposition calculation using by the result of margin algorithm
And obtain, formula is as follows:
ScaledResource=originalResource* (1+multiplier/confidence) ^exponent;
Wherein, originalResource is the result TargetResource of aforementioned margin algorithm;Confidence is to adopt
The data duration of collection;The usage amount upper limit and the multiplier value of usage amount lower limit are respectively 1 and 0.001;The usage amount upper limit and
The exponent value of usage amount lower limit is respectively 1 and -2;
S23, the target resource usage amount, the usage amount upper limit and usage amount lower numerical limit of the container of the recommendation of pod are updated into deposit
In status field, prepare for down-stream.
3. a kind of method based on the resource dynamic adjustment of kubernetes container according to claim 2, it is characterised in that:
The step S3 specifically includes the following contents:
S31, the resource situation information that a pod is obtained every the t1 time;
S32, the event information that pod is obtained by the status field information of pod, when the presetting t2 time started in pod
Occurred to then follow the steps S35 when memory uses up the event for being forced to delete in section, it is no to then follow the steps S33;
S33, configuration resource usage amount information is obtained by the resource field information of pod, when the configuration resource usage amount of pod
S35 is thened follow the steps when not between the aforementioned usage amount upper limit and usage amount lower limit, it is no to then follow the steps S34;
S34, the configuration resource usage amount of pod and preceding aim resource usage amount are compared, when configuration resource usage amount with
The gap ratio of target resource usage amount then terminates process and exits when being less than presetting numerical value k, no to then follow the steps S35;
S35, pod is carried out to chase operation, executes pod reboot operation.
4. a kind of method based on the resource dynamic adjustment of kubernetes container according to claim 3, it is characterised in that:
In the step S3, t1 is configurable value, and default value is 1 minute;T2 is configurable value, and default value is 10 minutes;K is can
The value of configuration, default value 10%.
5. a kind of method based on the resource dynamic adjustment of kubernetes container according to claim 1, it is characterised in that:
The particular content of the step S4 is as follows:
The new pod that creates that S41, interception are sent to kubernetes application programming interface server is requested;
S42, it updates target resource usage amount as the request usage amount of pod resource into the resource distribution of pod, and
Modified pod resource allocation information is sent to kubernetes application programming interface server, completes the wound of new pod
It builds.
6. a kind of vertical automatic telescopic device of Pod container group, it is characterised in that: including the following contents:
Acquisition module: for passing through the index collection of server built in kubernetes to the real-time of pod every preheating setting time
Container real resource usage amount;And
Nominator's module: for being integrated, calculating separately and record pushing away for pod according to the real resource usage amount being collected into
The target resource usage amount for the container recommended;And
Renovator module: for persistently monitoring the resource situation information of pod, when the target resource usage amount of pod is higher or lower than
When container configures resource usage amount presetting numerical value, or when memory occur within the presetting period using up to be forced to delete
When the event removed, pod is restarted;And
Admission controller module: it is arrived for intercepting request to create when pod is restarted, and by the update of preceding aim resource usage amount
In the resource distribution of pod, and new pod is created with the configuration.
7. a kind of vertical automatic telescopic device of Pod container group according to claim 6, it is characterised in that: nominator's mould
The particular content of block is as follows:
For obtaining the service condition information including CPU and memory usage amount of each pod in cluster, and integrated;With
And
Using margin algorithm and confidence algorithm, target resource usage amount, the usage amount upper limit and usage amount are respectively obtained
Lower limit;It is specific as follows:
Target resource usage amount is calculated using margin algorithm, and margin algorithmic formula is as follows:
TargetResource=RecourceAmount* (1+marginFraction), wherein RecourceAmount is to receive
The real resource usage amount collected, marginFraction are what target resource resource usage amount had more than real resource usage amount
Multiple;
The usage amount upper limit and usage amount lower limit all substitute into confidence algorithm superposition calculation using by the result of margin algorithm
And it obtains;Confidence algorithmic formula is as follows:
ScaledResource=originalResource* (1+multiplier/confidence) ^exponent;
Wherein, originalResource is the result TargetResource of aforementioned margin algorithm;Confidence is to adopt
The data duration of collection;The usage amount upper limit and the multiplier value of usage amount lower limit are respectively 1 and 0.001;The usage amount upper limit and
The exponent value of usage amount lower limit is respectively 1 and -2;And
The target resource usage amount, the usage amount upper limit and usage amount lower numerical limit of the container of the recommendation of pod are updated into deposit
In status field, prepare for down-stream.
8. a kind of vertical automatic telescopic device of Pod container group according to claim 7, it is characterised in that: the renovator mould
The particular content of block is as follows:
For obtaining the resource situation information of a pod every the t1 time;And
The event information that pod is obtained by the status field information of pod, when within the presetting t2 period that pod starts
Occurred memory use up be forced delete event when, then by pod carry out expulsion execute pod reboot operation;And it is if presetting
The t2 period in do not occur memory use up be forced delete event then further include the following contents,
Configuration resource usage amount information is obtained by the resource field information of pod, when the configuration resource usage amount of pod does not exist
When between the aforementioned usage amount upper limit and usage amount lower limit, then pod is subjected to expulsion and executes pod reboot operation;And if configuration money
Source usage amount then further includes the following contents between the aforementioned usage amount upper limit and usage amount lower limit,
The configuration resource usage amount of pod and preceding aim resource usage amount are compared, when configuration resource usage amount and target
The gap ratio of resource usage amount then terminates process and exits when being less than presetting numerical value k, and pod is otherwise carried out expulsion execution
Pod reboot operation.
9. a kind of vertical automatic telescopic device of Pod container group according to claim 8, it is characterised in that: the renovator mould
In block, t1 is configurable value, and default value is 1 minute;T2 is configurable value, and default value is 10 minutes;K is configurable
Value, default value 10%.
10. a kind of vertical automatic telescopic device of the Pod container group according to claim 5 or 7, it is characterised in that: the access
The particular content of controller module is as follows:
For intercepting the new pod request of creation for being sent to kubernetes application programming interface server;And
It updates target resource usage amount as the request usage amount of pod resource into the resource distribution of pod, and will repair
Pod resource allocation information after changing is sent to kubernetes application programming interface server, completes the creation of new pod.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910566915.1A CN110287029A (en) | 2019-06-27 | 2019-06-27 | A method of it is adjusted based on kubernetes container resource dynamic |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910566915.1A CN110287029A (en) | 2019-06-27 | 2019-06-27 | A method of it is adjusted based on kubernetes container resource dynamic |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110287029A true CN110287029A (en) | 2019-09-27 |
Family
ID=68007754
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910566915.1A Pending CN110287029A (en) | 2019-06-27 | 2019-06-27 | A method of it is adjusted based on kubernetes container resource dynamic |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110287029A (en) |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110990121A (en) * | 2019-11-28 | 2020-04-10 | 中国—东盟信息港股份有限公司 | Kubernetes scheduling strategy based on application portrait |
CN111158908A (en) * | 2019-12-27 | 2020-05-15 | 重庆紫光华山智安科技有限公司 | Kubernetes-based scheduling method and device for improving GPU utilization rate |
CN111290768A (en) * | 2020-01-22 | 2020-06-16 | 北京百度网讯科技有限公司 | Updating method, device, equipment and medium for containerization application system |
CN111506412A (en) * | 2020-04-22 | 2020-08-07 | 上海德拓信息技术股份有限公司 | Distributed asynchronous task construction and scheduling system and method based on Airflow |
CN111708609A (en) * | 2020-06-19 | 2020-09-25 | 中国—东盟信息港股份有限公司 | Kubernetes container based implementation method and system for configuring dictionary and security dictionary |
CN112000422A (en) * | 2020-07-17 | 2020-11-27 | 苏州浪潮智能科技有限公司 | Method and device for preventing POD memory overflow in container arrangement frame |
CN112015433A (en) * | 2020-08-28 | 2020-12-01 | 北京浪潮数据技术有限公司 | Resource scheduling method and device, electronic equipment and storage medium |
CN112099911A (en) * | 2020-08-28 | 2020-12-18 | 中国—东盟信息港股份有限公司 | Method for constructing dynamic resource access controller based on Kubernetes |
CN112181603A (en) * | 2020-10-26 | 2021-01-05 | 浪潮云信息技术股份公司 | Controllable vertical automatic retractor implementation method and system in cloud environment |
CN112181597A (en) * | 2020-10-12 | 2021-01-05 | 成都精灵云科技有限公司 | Dynamic updating method for use limit of container resource |
CN112506444A (en) * | 2020-12-28 | 2021-03-16 | 南方电网深圳数字电网研究院有限公司 | Kubernetes cluster-based expansion and contraction capacity control method and device and electronic equipment |
CN112559186A (en) * | 2020-12-22 | 2021-03-26 | 北京云思畅想科技有限公司 | Novel Kubernetes container resource expansion and contraction method |
CN112783642A (en) * | 2019-11-11 | 2021-05-11 | 阿里巴巴集团控股有限公司 | In-container logic configuration method, device and computer readable medium |
CN112783608A (en) * | 2021-01-29 | 2021-05-11 | 上海哔哩哔哩科技有限公司 | Method and device for adjusting container resources in container cluster |
WO2021103790A1 (en) * | 2019-11-26 | 2021-06-03 | 北京京东尚科信息技术有限公司 | Container scheduling method and apparatus, and non-volatile computer-readable storage medium |
WO2021103646A1 (en) * | 2019-11-26 | 2021-06-03 | 华为技术有限公司 | Pod deployment method and device |
CN112925695A (en) * | 2021-03-29 | 2021-06-08 | 浪潮云信息技术股份公司 | Method for automatically updating configuration file for configuring fluent |
CN113342461A (en) * | 2021-05-31 | 2021-09-03 | 北京市商汤科技开发有限公司 | Equipment mounting method and device, computer equipment and readable storage medium |
CN113590415A (en) * | 2021-06-30 | 2021-11-02 | 郑州云海信息技术有限公司 | Port management system, method, device and medium of deep learning training platform |
CN113872997A (en) * | 2020-06-30 | 2021-12-31 | 华为技术有限公司 | Container group POD reconstruction method based on container cluster service and related equipment |
CN113938379A (en) * | 2021-09-29 | 2022-01-14 | 浪潮云信息技术股份公司 | Method for dynamically loading cloud platform log acquisition configuration |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109144727A (en) * | 2018-08-21 | 2019-01-04 | 郑州云海信息技术有限公司 | The management method and device of resource in cloud data system |
CN109783218A (en) * | 2019-01-24 | 2019-05-21 | 中国—东盟信息港股份有限公司 | A kind of container dispatching method based on Kubernetes container cluster with time correlation connection |
-
2019
- 2019-06-27 CN CN201910566915.1A patent/CN110287029A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109144727A (en) * | 2018-08-21 | 2019-01-04 | 郑州云海信息技术有限公司 | The management method and device of resource in cloud data system |
CN109783218A (en) * | 2019-01-24 | 2019-05-21 | 中国—东盟信息港股份有限公司 | A kind of container dispatching method based on Kubernetes container cluster with time correlation connection |
Non-Patent Citations (2)
Title |
---|
TOADER SEBASTIAN: "vertical Pod autoscaling", 《HTTPS://BANZAICLOUD.COM/BLOG/K8S-VERTICAL-POD-AUTOSCALER/》 * |
WARMCHANG 等: "Vertical Pod Autoscaler", 《HTTPS://GITHUB.COM/KUBERNETES/COMMUNITY/BLOB/MASTER/CONTRIBUTORS/DESIGN-PROPOSALS/AUTOSCALING/VERTICAL-POD-AUTOSCALER.MD》 * |
Cited By (29)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112783642A (en) * | 2019-11-11 | 2021-05-11 | 阿里巴巴集团控股有限公司 | In-container logic configuration method, device and computer readable medium |
WO2021103646A1 (en) * | 2019-11-26 | 2021-06-03 | 华为技术有限公司 | Pod deployment method and device |
WO2021103790A1 (en) * | 2019-11-26 | 2021-06-03 | 北京京东尚科信息技术有限公司 | Container scheduling method and apparatus, and non-volatile computer-readable storage medium |
CN110990121A (en) * | 2019-11-28 | 2020-04-10 | 中国—东盟信息港股份有限公司 | Kubernetes scheduling strategy based on application portrait |
CN111158908A (en) * | 2019-12-27 | 2020-05-15 | 重庆紫光华山智安科技有限公司 | Kubernetes-based scheduling method and device for improving GPU utilization rate |
CN111290768A (en) * | 2020-01-22 | 2020-06-16 | 北京百度网讯科技有限公司 | Updating method, device, equipment and medium for containerization application system |
CN111290768B (en) * | 2020-01-22 | 2023-10-20 | 北京百度网讯科技有限公司 | Updating method, device, equipment and medium of containerized application system |
CN111506412A (en) * | 2020-04-22 | 2020-08-07 | 上海德拓信息技术股份有限公司 | Distributed asynchronous task construction and scheduling system and method based on Airflow |
CN111506412B (en) * | 2020-04-22 | 2023-04-25 | 上海德拓信息技术股份有限公司 | Airflow-based distributed asynchronous task construction and scheduling system and method |
CN111708609A (en) * | 2020-06-19 | 2020-09-25 | 中国—东盟信息港股份有限公司 | Kubernetes container based implementation method and system for configuring dictionary and security dictionary |
CN113872997B (en) * | 2020-06-30 | 2022-08-26 | 华为技术有限公司 | Container group POD reconstruction method based on container cluster service and related equipment |
CN113872997A (en) * | 2020-06-30 | 2021-12-31 | 华为技术有限公司 | Container group POD reconstruction method based on container cluster service and related equipment |
CN112000422B (en) * | 2020-07-17 | 2022-08-05 | 苏州浪潮智能科技有限公司 | Method and device for preventing POD memory overflow in container arrangement frame |
CN112000422A (en) * | 2020-07-17 | 2020-11-27 | 苏州浪潮智能科技有限公司 | Method and device for preventing POD memory overflow in container arrangement frame |
CN112015433A (en) * | 2020-08-28 | 2020-12-01 | 北京浪潮数据技术有限公司 | Resource scheduling method and device, electronic equipment and storage medium |
CN112099911B (en) * | 2020-08-28 | 2024-02-13 | 中国—东盟信息港股份有限公司 | Method for constructing dynamic resource access controller based on Kubernetes |
CN112099911A (en) * | 2020-08-28 | 2020-12-18 | 中国—东盟信息港股份有限公司 | Method for constructing dynamic resource access controller based on Kubernetes |
CN112181597A (en) * | 2020-10-12 | 2021-01-05 | 成都精灵云科技有限公司 | Dynamic updating method for use limit of container resource |
CN112181597B (en) * | 2020-10-12 | 2024-01-19 | 成都精灵云科技有限公司 | Dynamic updating method for container resource use limit |
CN112181603A (en) * | 2020-10-26 | 2021-01-05 | 浪潮云信息技术股份公司 | Controllable vertical automatic retractor implementation method and system in cloud environment |
CN112559186A (en) * | 2020-12-22 | 2021-03-26 | 北京云思畅想科技有限公司 | Novel Kubernetes container resource expansion and contraction method |
CN112506444A (en) * | 2020-12-28 | 2021-03-16 | 南方电网深圳数字电网研究院有限公司 | Kubernetes cluster-based expansion and contraction capacity control method and device and electronic equipment |
CN112783608A (en) * | 2021-01-29 | 2021-05-11 | 上海哔哩哔哩科技有限公司 | Method and device for adjusting container resources in container cluster |
CN112925695A (en) * | 2021-03-29 | 2021-06-08 | 浪潮云信息技术股份公司 | Method for automatically updating configuration file for configuring fluent |
CN113342461A (en) * | 2021-05-31 | 2021-09-03 | 北京市商汤科技开发有限公司 | Equipment mounting method and device, computer equipment and readable storage medium |
CN113342461B (en) * | 2021-05-31 | 2023-04-07 | 北京市商汤科技开发有限公司 | Equipment mounting method and device, computer equipment and readable storage medium |
CN113590415B (en) * | 2021-06-30 | 2023-09-22 | 郑州云海信息技术有限公司 | Port management system, method, equipment and medium of deep learning training platform |
CN113590415A (en) * | 2021-06-30 | 2021-11-02 | 郑州云海信息技术有限公司 | Port management system, method, device and medium of deep learning training platform |
CN113938379A (en) * | 2021-09-29 | 2022-01-14 | 浪潮云信息技术股份公司 | Method for dynamically loading cloud platform log acquisition configuration |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110287029A (en) | A method of it is adjusted based on kubernetes container resource dynamic | |
CN105049268B (en) | Distributed computing resource distribution system and task processing method | |
US9112782B2 (en) | Reactive auto-scaling of capacity | |
CN104579761B (en) | A kind of nosql clusters automatic configuration system and method for automatic configuration based on cloud computing | |
CN107580023A (en) | A kind of the stream process job scheduling method and system of dynamic adjustment task distribution | |
CN105338027B (en) | Carry out the method, system and device of video data cloud storage | |
CN110659741A (en) | AI model training system and method based on piece-splitting automatic learning | |
CN110413391A (en) | Deep learning task service method for ensuring quality and system based on container cluster | |
US9058428B1 (en) | Software testing using shadow requests | |
CN111399986A (en) | Method and device for configuring Pod resource quota | |
CN102622303A (en) | Memory overload control method and device | |
CN112445575B (en) | Multi-cluster resource scheduling method, device and system | |
CN113010260A (en) | Elastic expansion method and system for container quantity | |
CN112685153A (en) | Micro-service scheduling method and device and electronic equipment | |
CN103701934A (en) | Resource optimal scheduling method and virtual machine host machine optimal selection method | |
CN114035941A (en) | Resource scheduling system, method and computing device | |
CN107291536A (en) | Application task stream scheduling method under a kind of cloud computing environment | |
CN112365366A (en) | Micro-grid management method and system based on intelligent 5G slice | |
CN107967175A (en) | A kind of resource scheduling system and method based on multiple-objection optimization | |
CN109936471B (en) | Multi-cluster resource allocation method and device | |
CN106850330A (en) | Intelligentized cloud desktop Performance Test System and method | |
CN102026228A (en) | Statistical method and equipment for communication network performance data | |
CN113568759B (en) | Cloud computing-based big data processing method and system | |
CN107479974A (en) | A kind of dispatching method of virtual machine and device | |
CN101212342B (en) | Multi-version network element network management method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190927 |