Details
-
Bug
-
Resolution: Not a Bug
-
Critical
-
7.6.0
-
7.6.0-1824
-
Untriaged
-
0
-
Unknown
Description
The automated test has failed with shard limit crossed error -
The test does the following -
Sets flush buffer quota to 0.5 MB.
Sets maxDiskUsagePerShard to 26843545 ( ~ 26 MB).
Start creating indexes in batches of 10 ( Around 60 indexes in total).
During each iteration, fetch the physical shards created. (By querying /getIndexMetadata) and calculate if it has exceeded the limit.
Limit calculations are as follows -
max_shard_count = settings['indexer.plasma.shardLimitPerTenant'] |
flush_buffer_quota = settings['indexer.plasma.flushBufferQuota'] |
mem_quota = self.get_indexer_mem_quota()
|
shard_limit = math.ceil(mem_quota * 0.9 * flush_buffer_quota / 100) |
if shard_limit % 2 != 0: |
shard_limit += 1 |
Now at some point, the test has ended up with 6 shards
723809672567083874, 12326645128606720287, 15826462576033473764, 15233062409255651751, 183416447645531922, 12205676462534385018] |
This had passed a few times after the fix in https://issues.couchbase.com/browse/MB-59465, but seems to have failed again.
Not really sure if the maxDiskUsagePerShard setting has any effect, so I can change the test depending upon RCA.
s3://cb-customers-secure/shardlimitexceeded/2023-11-21/shardlimitexceeded.zip
Assigning this to Varun Velamuri (because of https://issues.couchbase.com/browse/MB-59465) . Please re-assign if necessary.