[
https://issues.apache.org/jira/browse/HBASE-16095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15369222#comment-15369222
]
Andrew Purtell commented on HBASE-16095:
----------------------------------------
bq. I feel that the priority region opening might still be useful for other
contexts as well (like opening framework level tables sooner), hence we should
still pursue this.
This makes sense IMHO.
This is not per request priorities (HBASE-15816), and we may have an
(orthogonal) use case for those someday, but it's also possible this change
alone is useful enough to Phoenix: Assign index tables a higher priority than
the primary table and that may solve their deadlock challenges.
Not thrilled about the new event type but alternatives seem worse IMHO
{code}
M_RS_OPEN_PRIORITY_REGION (26, ExecutorType.RS_OPEN_PRIORITY_REGION),
{code}
We handle priority opens of META with its own pool, and will have another
priority pool after this change. The two priority pools aren't aware of each
other and may execute concurrently.
In general we are approaching this in an ad hoc manner with static pools,
predefined and limited QoS levels, and magic constants. Why not have a dynamic
dispatch. Allocate a handler pool at every distinct priority level supplied by
htd.getPriority(). Such pools could have only one core thread and a
configurable upper bound on pool size, with the additional workers terminating
after like one minute of inactivity. Use the same approach for dispatch to
META: META handlers would become just a pool allocated at the highest priority
level. The downside is more work at dispatch than simple test-and-branch with
precompiled constants. Just a thought.
{code}
if (htd.getPriority() >= HConstants.ADMIN_QOS) {
regionServer.service.submit(new OpenPriorityRegionHandler(
regionServer, regionServer, region, htd,
masterSystemTime));
} else {
regionServer.service.submit(new OpenRegionHandler(
regionServer, regionServer, region, htd,
masterSystemTime));
}
}
{code}
> Add priority to TableDescriptor and priority region open thread pool
> --------------------------------------------------------------------
>
> Key: HBASE-16095
> URL: https://issues.apache.org/jira/browse/HBASE-16095
> Project: HBase
> Issue Type: Bug
> Reporter: Enis Soztutar
> Assignee: Enis Soztutar
> Fix For: 2.0.0, 1.4.0
>
> Attachments: hbase-16095_v0.patch, hbase-16095_v1.patch,
> hbase-16095_v2.patch
>
>
> This is in the similar area with HBASE-15816, and also required with the
> current secondary indexing for Phoenix.
> The problem with P secondary indexes is that data table regions depend on
> index regions to be able to make progress. Possible distributed deadlocks can
> be prevented via custom RpcScheduler + RpcController configuration via
> HBASE-11048 and PHOENIX-938. However, region opening also has the same
> deadlock situation, because data region open has to replay the WAL edits to
> the index regions. There is only 1 thread pool to open regions with 3 workers
> by default. So if the cluster is recovering / restarting from scratch, the
> deadlock happens because some index regions cannot be opened due to them
> being in the same queue waiting for data regions to open (which waits for
> RPC'ing to index regions which is not open). This is reproduced in almost all
> Phoenix secondary index clusters (mutable table w/o transactions) that we
> see.
> The proposal is to have a "high priority" region opening thread pool, and
> have the HTD carry the relative priority of a table. This maybe useful for
> other "framework" level tables from Phoenix, Tephra, Trafodian, etc if they
> want some specific tables to become online faster.
> As a follow up patch, we can also take a look at how this priority
> information can be used by the rpc scheduler on the server side or rpc
> controller on the client side, so that we do not have to set priorities
> manually per-operation.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
