Adam Rempter created ATLAS-3254:
-----------------------------------
Summary: Atlas entity with large array of refs causes performance
issues for lineage
Key: ATLAS-3254
URL: https://issues.apache.org/jira/browse/ATLAS-3254
Project: Atlas
Issue Type: Improvement
Components: atlas-core, atlas-webui
Affects Versions: 2.0.0, 1.0.0
Reporter: Adam Rempter
We use “aws_s3_pseudo_dir” type from 3020-aws_s3_typedefs.json model.
It has following property:
"name": "s3Objects",
"typeName": "array<aws_s3_object>"
Now in AWS buckets you can have thousands of objects. This causes that
s3Objects array grows quite quickly, causing aws_s3_pseudo_dir entity Json to
rich easly few MBs.
Then we start seeing problems like:
* UI is dying on displaying entity properties or lineage
* Error in logs: audit record too long: entityType=aws_s3_pseudo_dir,
guid=24398271-6ba0-4db5-adfa-38e432dc55ce, size=1053931; maxSize=1048576.
entity attribute values not stored in audit (EntityAuditListenerV2:234)
* Some errors with write to HBase (java.lang.IllegalArgumentException:
KeyValue size too large, as workaround we set hbase.client.keyvalue.maxsize
param to 0)
* kafka consumer errors (we can of course set some parameters on consumer, but
I think it is just workaround)
…
Exception in NotificationHookConsumer (NotificationHookConsumer:332)
org.apache.kafka.clients.consumer.CommitFailedException: Commit cannot be
completed since the group has already rebalanced and assigned the partitions to
another member. This means that the time between subsequen
t calls to poll() was longer than the configured max.poll.interval.ms, which
typically implies that the poll loop is spending too much time message
processing. You can address this either by increasing the sessio
n timeout or by reducing the maximum size of batches returned in poll() with
max.poll.records.
…
Specifying pseudo_dir is required for s3objects:
name": "pseudoDirectory",
"typeName": "aws_s3_pseudo_dir",
"cardinality": "SINGLE",
"isIndexable": false,
*"isOptional": false,*
"isUnique": false,
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
