[
https://issues.apache.org/jira/browse/HBASE-10616?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13913643#comment-13913643
]
Nick Dimiduk commented on HBASE-10616:
--------------------------------------
{noformat}
//TODO: should we record each GET in the batch as one READ
{noformat}
This should be necessary if you want to directly compare different runs (ie, no
batching, batchsize=10, batchsize=50, &c.), so yes, I think it's desirable.
OTOH, I don't have this in PerfEval yet, though I believe the reported
throughput metrics are correct for this use.
{noformat}
+ // A sanity check for the key range.
+ if (k < startKey || k >= endKey) {
+ numReadErrors.incrementAndGet();
+ throw new AssertionError("Load tester logic error: proposed key " +
+ "to read " + k + " is out of range (startKey=" + startKey +
+ ", endKey=" + endKey + ")");
+ }
{noformat}
This appears to throw when a key falls outside of the target row range. Above
that, the {{keys[]}} is populated up to target batchsize ({{do {} while
(isMultiGet && numIter < batchSize);}}). Does that mean this assertion fails in
the event that the batch boundary and key range are not precisely aligned? I'd
expect this to happen at the end of every range unless you have other
calculations to enforce this alignment. PerfEval handles this case by checking
for any remaining buffered items and sending that partial batch at test
conclusion.
> Integration test for multi-get calls
> ------------------------------------
>
> Key: HBASE-10616
> URL: https://issues.apache.org/jira/browse/HBASE-10616
> Project: HBase
> Issue Type: Sub-task
> Reporter: Devaraj Das
> Assignee: Devaraj Das
> Attachments: 10616-1.txt
>
>
> HBASE-10572 adds a test that does 'get' from client. We should add a similar
> test for batch gets.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
