[
https://issues.apache.org/jira/browse/HBASE-9987?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13849939#comment-13849939
]
stack commented on HBASE-9987:
------------------------------
Is this ok?
+ Map<byte[], HRegionLocation> tableLocations =
getTableLocations(tableName);
+ // start to examine the cache. we can only do cache actions
+ // if there's something in the cache for this table.
+ rl = getCachedLocation(tableName, row);
+ if (rl != null) {
+ tableLocations.remove(rl.getRegionInfo().getStartKey());
}
Could another thread have set into tableLocations a good location and we are
removing it here? Previous it couldn't have happened because there was the
sync block around the whole operation.
Ditto here:
+ result = this.cachedRegionLocations.get(tableName);
+ // if tableLocations for this table isn't built yet, make one
+ if (result == null) {
+ result = new ConcurrentSkipListMap<byte[],
HRegionLocation>(Bytes.BYTES_COMPARATOR);
+ ConcurrentSkipListMap<byte[], HRegionLocation> old =
+ this.cachedRegionLocations.putIfAbsent(tableName, result);
+ if (old != null) {
+ return old;
Could another thread have inserted a result with perhaps some values in it?
I suppose it is not the end of the world if this happens.... We'll just miss
the cache a bit more often.
Could this be a problem?
- synchronized(this.cachedRegionLocations) {
- this.cachedRegionLocations.clear();
- this.cachedServers.clear();
- }
+ this.cachedRegionLocations.clear();
+ this.cachedServers.clear();
Could we give a bad location out if cachedRegionLocations is cleared but
cachedServers is not?
There seems to be a bit of dodgy updating now we no longer have sync blocks.
but perhaps it is ok; we just miss cache a few times more for a better perf
overall?
> Remove some synchronisation points in HConnectionManager
> --------------------------------------------------------
>
> Key: HBASE-9987
> URL: https://issues.apache.org/jira/browse/HBASE-9987
> Project: HBase
> Issue Type: Bug
> Components: Client
> Affects Versions: 0.98.0, 0.96.0
> Reporter: Nicolas Liochon
> Assignee: Nicolas Liochon
> Fix For: 0.98.0, 0.96.1
>
> Attachments: 9987.v1.patch, 9987.v2.patch
>
>
> Change a Map to a concurrentMap
> Removed the "cachedServer (introduced in HBASE-4785). I suspect that this
> function is not needed anymore as we also have a list of dead servers, and
> accessing the list is not blocking. I will dig into this more however.
> The patch gives a 10% improvement with the NoClusterClient.
--
This message was sent by Atlassian JIRA
(v6.1.4#6159)
