hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nicolas Liochon (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-10701) Cache invalidation improvements from client side
Date Wed, 12 Mar 2014 20:30:43 GMT

    [ https://issues.apache.org/jira/browse/HBASE-10701?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13932309#comment-13932309
] 

Nicolas Liochon commented on HBASE-10701:
-----------------------------------------

bq. Bytes.toString(regionName));
Should be a "Bytes.toStringBinary"

bq. RegionLocations
Seems ok to me.

bq. MetaReader
{code}
    } catch (Exception parseEx) {
      // Ignore. This is used with tableName passed as regionName.
    }
{code}
This is scary. We're wild on catching here, and this denotes a design issue (may be an old
one?)

bq. AsyncProcess / ConnectionManager
Are we sure that  locateRegionInMeta will not return a location with a null serverName?
AsyncProcess doesn't retry if the location is null because it expects that there is already
a retry there.

bq. RpcRetryingCallerWithReadReplicas 
??long start = System.nanoTime();??
This is not used

??if (exceptions == null) exceptions = new ArrayList<ExecutionException>(rl.size());??
the condition is always true here

??Result result = null;??
The scope could be narrowed.

??// the primary call failed with RetriesExhaustedException or DoNotRetryIOException??
bq. In case of a close connection
This should not be the case. This is something we should retry (and we do iirc). But may be
you were speaking about another close?
Note that I'm not against waiting until the bitter end. But if there is a coding/logic error
by the client/customer, (usually caught by DoNotRetryIOException), we will go on all servers
instead of only once, that's why I preferred the first approach in the original implementation.
DoNotRetryIOException means "logic error" usually...




> 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
>
>
> 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
(v6.2#6252)

Mime
View raw message