TKE experience upgrade: 24 tips to get started with K8s faster

Author

Wang Xiaowei, Tencent Cloud Container Product Manager, is keen to provide customers with an efficient way to use Kubernetes and provide customers with extreme cost reduction and efficiency services.

background

"Function" solves the problem of whether the product has or does not have a capability. After having the "function", how to help users to use the product more "smoothly" through good guidance, prompts, and interaction is a question worth thinking about.

For example: Lego already has more than 3700 different modules, these modules can spell an unlimited number of models. However, the presentation of different models is completely different, and it is possible that just a change in the position of a few small modules can bring unexpected effects.

After several years of hard work, TKE has put forward a lot of new features at the functional level: for example: through LB direct connection Pod to improve performance , cluster in-situ upgrade , virtual node , dynamic scheduler more Refer to Container Service Product Dynamic .

In addition, TKE also attaches importance to the overall user experience of users. For this reason, it visited a number of companies, collected a lot of needs, and carried out optimization and tool development based on the user’s attention to stratification. Users can use these optimization tips to make it smoother. Experience the container service smoothly, lower the barriers to product use and enhance the product use experience more "groundedly".

Boost list

Cluster/node related

Recommendation: Enhanced node search capabilities

Node search supports multi-dimensional search Label/IP/blocked/node status

Label search format is the same as native Kubernetes: "k1=v1,k2=v2"
Only accurate query is supported for IP multi-value search, and fuzzy query is supported for single IP search

use: node management -> node -> search or node pool details page -> drop down to the node list -> search

Recommendation: Optimization of node initialization error display

[Detailed display of errors related to CVM delivery failure] If the node purchase fails due to CVM delivery failure, in addition to reporting LAUNCH_FAILED (that is, delivery failure), the detailed reason for the delivery failure will also be displayed: for example, VPC subnet IP is insufficient/ Insufficient disk resources / insufficient balance / insufficient placement group quota and so on.
[Detailed display of errors related to node initialization] The steps of node initialization failure and corresponding information will be displayed in the node list in detail: for example, failed to hang disk, core component abnormal etc.

use: Node Management -> Node -> Click View Status in the status bar to view the error

Standardization of node pool node operation: support for blockade/eviction

The node operation in the node pool supports blockade/eviction, which has been fully synchronized with the operation of the node list in the cluster. You can efficiently manage the nodes in the node pool.

usage: node pool details page -> drop down to the node list -> operate on a single node in the operation column or select batch nodes and perform batch operations (click the [More Operations] button above to batch block/unblock nodes)

Delete protection logic optimization

Customer feedback: If the accidental deletion option is enabled when the cluster is created, the customer cannot delete it if the creation fails after enabling it.

Now you can turn off the delete protection function for the cluster in the [Cluster List Page].

Usage: cluster management -> cluster operation list -> more -> disable cluster delete protection

Public IP display optimization

Some customers reported that the configuration entry of the public network IP was too concealed, and the default bandwidth value was too small to meet business needs.

We take a graphical form (drag bar) to replace the input box to enhance the display effect.

How to use: new node/new node pool -> select model -> drag and drop to the bottom -> set public network bandwidth

Node pool display information structured / advanced parameter display

Add an overview of the core definitions/functions of the node pool, allowing customers to quickly understand the usage scenarios of the node pool

Structured display of node pool information (aggregated and displayed by modules, and more advanced parameters are displayed at the same time, which improves observability)
- Basic information of the node pool (information related to elastic scaling, number of nodes, etc.)
- Node configuration details (node template related configuration, configuration automatically inherited by node pool expansion nodes)

Node jump directly to CVM optimization

The old version of the Jump CVM button is unstable, and the entrance is not prominent.

use: hover the mouse to the node ID (if you directly click on the node ID, it will jump to the Node details page) -> click "jump to the CVM instance details page" -> it will jump to the CVM instance details page .

Added container network IP verification when creating nodes

Before creating a node, you will determine whether the container network can meet the requirements for the creation of this batch of nodes based on your container network configuration (network mode, network segment, etc.). Considering that you may not need to verify the container network in some special scenarios, we provide the parameter skip to allow you to skip the verification. (For detailed description of the parameters, please refer to API document )

use: new node/add existing node/new node pool -> more settings -> set skip parameter

Support to fill in the mount partition/LVM name when adding an existing node to the cluster

When adding an existing node, if you partition/LVM the disk, please fill in the partition name/LVM name you want to mount in the mount partition, If the disk is not partitioned or LVM, you don’t need to fill in the mount partition, just fill in Mount point .

use: add an existing node -> select a node -> check the data disk mount -> fill in the mount partition and mount point

Kubernetes object related

Recommendation: Batch import key-value pairs: configuration management & environment variable optimization

Support importing batches of key-value pairs from the clipboard at one time. When the variable name is empty, paste one or more lines of key=value or key: value key-value pairs in the variable name to realize fast batch input.

Usage 1: Workload -> New Load -> Instance Content

Usage 2: Configuration Management -> New ConfigMap/Secret -> Content

Recommendation: CronJob supports reserved quantity/pause/manual trigger/display of job list generated by Cronjob

number of successful reserved jobs : Prevent unlimited job generation and affect cluster performance
Number of failed reservation jobs : Prevent unlimited job generation and affect cluster performance
Pause : Cronjob will be stopped and no new jobs will be generated. supports resume operation
Manual trigger : No need to wait for the Crontab time set by Cronjob to support the use of the Job template in Cronjob to generate a new Job

Usage: Workload -> Cronjob -> New -> Number of successful reserved jobs/number of failed reservations

Usage: Workload -> Cronjob list -> More -> Manual trigger/pause

Click on the name of CronJob in the figure above to view the list of running Jobs generated by CronJob:

HostAliases productization

HostAliases used to add entries to the /etc/hosts file of the Pod, which can cover the resolution of the host name at the Pod level. To increase the DNS configuration of the Pod.

Usage: Workload -> New Load -> Advanced Settings -> Host Alias

Lifecycle productization

lifecycle : The lifecycle event hook processing function of the container. A container can specify a handler for each event. When a container is started, Kubernetes will immediately send a postStart event; before the container is terminated, Kubernetes will send a preStop event.

Usage: Workload -> New Load -> Instance Content

Service defaults to "access only within the cluster"; Service type explanation adds corresponding English nouns

"Access only within the cluster" access method is in line with usage habits
The comparison of Chinese and English terminology reduces the user's understanding cost, which is in line with the native usage habits of Kubernetes

One-click "re-deployment" is supported on Deployment, StatefulSet and DaemonSet

Like Deployment, StatefulSet and DaemonSet now support one-click redeployment, eliminating the trouble of reconfiguration after deletion.

Usage: Workload -> Load List -> More -> Redeploy

Ingress certificate supports inheriting LB certificate

It supports the direct use of CLB's certificate to automatically create Ingress's HTTPS certificate, eliminating the cumbersome Secret creation process.

Usage: Service and routing -> Ingress -> New Ingress -> Listening port check Https:443 -> New key -> Select server certificate

Secret supports TLS type

In addition to the CLB certificate, you sometimes use your own certificate. To save the certificate, you need to use TLS type Secret

Usage 1: Configuration Management -> Secret -> New Secret -> TLS Certificate

Usage 2: Service and routing -> Ingress -> New Ingress -> Nginx load balancer -> Listening port check Https:443 -> New key

subPathExpr supports configuration on the console

subPathExpr , supports environment variable to a subpath in Volume

Usage: Create new load -> use data volume -> add environment variable -> use mount point -> select subPathExpr

The resources in kube-system, kube-node-lease, and kube-public support modification in the console

In the past, in order to prevent users from accidentally touching, the modification of resources in these namespaces was restricted in the console, but in fact, some users themselves have the resource management permissions of these namespaces, so the new version removes these restrictions in the console, and the logic of permissions is changed. Judging by the background, if you don't have the permission of the relevant namespace, an error will be reported if you modify the resource.

Usage: New load -> select namespace -> modify resource configuration

Usage: authorization management -> select any entry -> RBAC policy generator -> select sub-account -> select namespace/permission

Added multiple floating reminders

Added floating prompts to the field names of all resource list pages, which is more suitable for users with smaller screens

The log entry container name in the workload adds a floating prompt, which is more suitable for users with smaller screens

Operation and maintenance capabilities related

Recommendation: Log collection supports automatic extraction of regular expressions based on log samples

When the single-line fully regular and multi-line fully regular extraction modes are selected in the log collection rules, regular expressions can be automatically generated according to the user's own log samples, and the corresponding key-value pairs are automatically extracted, without the user having to learn and configure by himself Regular expression

Usage: 1. Cluster operation and maintenance -> 2. Log rule -> 3. New -> 4. Enter the name of the log collection rule -> 5. Select "Single line-Complete regular" for the extraction method -> 6. Enter the log sample -> 7. Click the regular expression to automatically generate -> 8. Move the cursor to select the string that needs to be regularized -> 9. Enter the key -> 10. Confirm

Recommendation: Log collection supports automatic verification based on log samples to determine whether logs can be extracted normally

For the regular expressions, separators, etc. filled in by the user, verification is supported according to the user log sample, and all values are automatically extracted to verify whether the log can be parsed normally, and to avoid finding an expression configuration error when retrieving the log.

Usage: 1. Cluster operation and maintenance -> 2. Log rule -> 3. New -> 4. Enter the name of the log collection rule -> 5. Select "Single line-Complete regular" for extraction method -> 6. Enter log sample -> 7. Manually enter the regular expression -> 8. Verify

Log collection supports multi-line regular extraction mode

Supports multi-line regular parsing mode (applicable to the requirement of retrieving logs with key-value pairs for multi-line logs such as java programs)

Note: To use the multi-line regular extraction mode, you need to contact the backend to upgrade the collection component version

use: Create a new log collection rule -> Select "Multi-line-Complete Regular Mode" in the second step of the extraction mode

Hosting Prometheus supports Singapore, Virginia and other regions

Cloud native monitoring (hosted Prometheus) is available in Singapore/Virginia and other regions, welcome to use.

to sum up

Kubernetes itself is powerful, but the modules are complex and the cost of getting started is high. TKE is committed to the upgrade of functions and experience. If you have any

[Tencent Cloud Native] Yunshuo new products, Yunyan new technology, Yunyou Xinhuo, Yunxiang information, scan the QR code to follow the public account of the same name, and get more dry goods in time! !