[
https://issues.apache.org/jira/browse/HBASE-2925?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12906002#action_12906002
]
stack commented on HBASE-2925:
------------------------------
@Robert ...I wonder if it defeats the purpose of having a connection cache from
the first place.
As I see it, the main benefit to the cache is saving on region lookups, setup
of zk connection, and master proxy setup. Having the likes of the following
config cached is secondary: e.g.
{code}
private final long pause;
private final int numRetries;
private final int maxRPCAttempts;
private final long rpcTimeout;
private final int prefetchRegionLimit;
{code}
In fact, I'm now thinking that if a user changes any of the above in a
Configuration that is being used as a key in HCM#HBASE_INSTANCES, that its ok
if we don't notice the change as long as we doc the fact that the Configuration
is read on instantiation of the HTable and then no more. I don't tthink this
should 'surprise' the user too much and they can just go create new HTable with
the new Configuration if they really want their new config. to take hold
(Making things work this way is similar to what you suggest only you would copy
the Configuration before adding it as a key).
As to your last suggestion, I think we should move away from trying to equate
Configurations at all; there be daemons that way.
Let me know what you think. If you are agreeable, I'll work up the patch some
more mostly adding doc clarifying what we've agreed here.
> LRU of HConnectionManager.HBASE_INSTANCES breaks if HBaseConfiguration is
> changed
> ---------------------------------------------------------------------------------
>
> Key: HBASE-2925
> URL: https://issues.apache.org/jira/browse/HBASE-2925
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.20.3, 0.90.0
> Reporter: Robert Mahfoud
> Attachments: 2925.txt, SimpleHConnectionManagerLeakReplicator.java
>
>
> {{HConnectionManager.getConnection(config)}} caches the created
> {{TableServer}} in {{HBASE_INSTANCES}} (a {{LinkedHashMap}} ) which is keyed
> by the configuration instance itself.
> Given the current implementation of {{hashCode()}} (and {{equals()}}) of
> {{HBaseConfiguration}}, the hash code of the configuration is changed if any
> of its properties are changed, which will cause the keys of
> {{HBASE_INSTANCES}} to be inconsistent with the hashtable that contains them,
> making some entries unreachable.
> In this case, when the map's LRU strategy needs to remove the oldest entry,
> it tries to remove it based on the oldest key, which no longer gives the
> original hash code, therefore the lookup in
> {{HBASE_INSTANCES.remove(oldest)}} doesn't actually remove anything.
> This has been observed to lead to OOM errors in long running clients.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
