ramkrish86 commented on a change in pull request #2483:
URL: https://github.com/apache/hbase/pull/2483#discussion_r497995964
##########
File path:
hbase-hadoop-compat/src/main/java/org/apache/hadoop/hbase/regionserver/MetricsRegionServerSource.java
##########
@@ -571,6 +571,9 @@
String RPC_SCAN_REQUEST_COUNT = "rpcScanRequestCount";
String RPC_SCAN_REQUEST_COUNT_DESC =
"Number of rpc scan requests this RegionServer has answered.";
+ String RPC_FULL_SCAN_REQUEST_COUNT = "rpcFullScanRequestCount";
+ String RPC_FULL_SCAN_REQUEST_COUNT_DESC =
+ "Number of rpc scan requests that were possible full region scans.";
Review comment:
`Looks good. Some nits in below. Main question is if this count has
value? If so, how? Thanks.`
@saintstack - the main reason was that we generally don't get an insight if
the user has issued a scan without really setting a start and stop row. Mostly
they want a specific set of rows in the scan but they generally miss setting
the start and stop row and depend on filters. Though filters are an option but
still deciding the rowkeys upfront to ensure we make use of the start and stop
row is a good practice. If we have a metric to capture we can atleast indicate
the pattern but whether it is really intentional or not may be something which
as experts we can decide and inform the user. Also though some time it is
intentional it might not be the actual business logic that is doing the scan
but a sanity check that the user is running to do some verifications. that is
not latency sensitive but still it will evict the cached data which is actually
needed for the real latency sensitive queries. In such cases atleast we can
suggest that the scan that is doing a full range/table scan c
an be made not to use the cache by setting `scan#setCaching(false)`
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
