hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Enis Soztutar (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-10701) Cache invalidation improvements from client side
Date Fri, 14 Mar 2014 06:12:44 GMT

     [ https://issues.apache.org/jira/browse/HBASE-10701?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Enis Soztutar updated HBASE-10701:

    Attachment: hbase-10701_v3.patch

Thanks Nicolas for the careful review. 

I've changed the patch so that I dropped the approach or using HRL's with null ServerNames.
Instead we still set the HRL item as null inside RegionLocations. RegionLocations now, can
contain null elements at the tail of the array as well. This enables the cache to know about
how many replicas there are, but the locations might still be unknown. 

I've been testing this with
hbase org.apache.hadoop.hbase.test.IntegrationTestTimeBoundedRequestsWithRegionReplicas -Dhbase.IntegrationTestTimeBoundedRequestsWithRegionReplicas.runtime=600000
-DIntegrationTestTimeBoundedRequestsWithRegionReplicas.num_write_threads=30 -DIntegrationTestTimeBoundedRequestsWithRegionReplicas.region_replication=3
-DIntegrationTestTimeBoundedRequestsWithRegionReplicas.num_read_threads=30 -Dhbase.ipc.client.allowsInterrupt=true
it seems the issues are fixed. However, I notice that the test most of the time dies with
OOM, cannot create native thread, because the number of threads grow unbounded (north of 4K).

Tried setting -Dhbase.hconnection.threads.max=512 with no results so far. 

One other issue (probably related) was that the RPC's would not start for a long time and
timeout the gets (10-20 secs) because the thread pool executor does not schedule the tasks
in the CompletionService from RpcRetryingCallerWithReadReplicas. Do you have any opinion around
this? Should we create a secondary pool for the backup requests? If we address the thread
growing problem, probably this will be fixed as well. 

The v3 patch also addresses your comments, except for the DoNotRetryEx. We'll have to get
this running consistently before addressing that I think. 

> Cache invalidation improvements from client side
> ------------------------------------------------
>                 Key: HBASE-10701
>                 URL: https://issues.apache.org/jira/browse/HBASE-10701
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Enis Soztutar
>            Assignee: Enis Soztutar
>             Fix For: hbase-10070
>         Attachments: hbase-10701_v1.patch, hbase-10701_v2.patch, hbase-10701_v3.patch
> Running the integration test in HBASE-10572, and HBASE-10355, it seems that we need some
changes for cache invalidation of meta entries from the client side in backup RPCs. 
> Mainly the RPC's made for replicas should not invalidate the cache for all the replicas
(for example on RegionMovedException, connection error etc). 

This message was sent by Atlassian JIRA

View raw message