A clique in the YTsaurus web interface

Cliques don't have their own web interface, but you can view all clique-related objects using the YTsaurus web interface.

Strawberry Node

The clique's configuration and persistent state are stored in Cypress at //sys/strawberry/chyt/<alias>.

The node description provides housekeeping information: whether the clique is running, in what pool, what the status of the corresponding YTsaurus operation is, and so on. This node also contains a YSON document called speclet, which stores all the clique settings set with the yt clickhouse ctl set-option command.

In addition, you can retrieve information from the strawberry node using the yt clickhouse ctl get-speclet and yt clickhouse ctl status commands in our CLI.

Access Control Object Node

The clique's ACL is stored in a specialized access control object node located at //sys/access_control_object_namespaces/chyt/<alias>. In this node, you can view all the accesses granted for the clique.

YT Operation

To access the web interface of the YTsaurus system operation used to start the clique instances, follow the link from the strawberry node description or the link from the yt clickhouse ctl status command output.

There is a lot of useful information on the operation page to help with the operation of CHYT.

Description

The Description section contains various system information about the clique and a number of links.

• ytserver-log-tailer, ytserver-clickhouse, clickhouse-trampoline — these sections describe versions of various components of the CHYT server code.

Specification

The Specification section contains information about how exactly the clique was started.

• Wrapper version: Contains the version of the launcher that started the clique.
• Command: Contains the clique start string.

The last two values are very useful if you need, for example, to restart a more recent version of the clique. You can either address the user who started it or reproduce the start command yourself.

Jobs

The Jobs section contains information about the jobs of the operation in which the clique was started. Remember that one job corresponds to one instance in a clique, that is, one ClickHouse server.

In this section, in terms of CHYT instances:

• total: The total number of instances in a clique.
• pending: The number of instances out of the total number that have not yet been started. A non-zero number in this section means that the clique does not have enough resources to keep all instances running at once — a bad situation for the clique.
• running: The number of running instances out of the total number, i.e. running + pending = total.
• completed: The number of gracefully preempted instances. This includes those instances that were subject to preemption and were able to complete all the queries running on them before they were forcibly aborted.
• failed: The number of failed instances. Instances can fail for various reasons and the most common are: OOM (out of memory) or bugs in the CHYT or ClickHouse code.
• aborted: The number of aborted instances. Some aborts happen due to insurmountable circumstances (for example, a node failed) and nothing can be done with them. If there are jobs aborted due to preemption, this is a bad sign that indicates a lack of resources.
• lost: Jobs must not occur in the clique.