[
https://issues.apache.org/jira/browse/ATLAS-4742?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jagadesh Kiran N updated ATLAS-4742:
------------------------------------
Description:
We have tried setting up Apache atlas(2.2.0) in High availability mode in
Kubernetes with separate pods for two instance.
When we simulate a failure on active instance, the failover from passive
instance from active instance is taking over an hour to complete the
transition. This happens when we have created around 300 entities whereas when
we simulated the same failover on fresh install of Atlas, it completed the
transition in 15 minutes.
This is the log before the failover for the passive instance:
The logs after the failover is in the attached file. This is the failover
during a fresh install. It is doing the similar process but taking more time
when there are more entities created in Atlas.
Warning : [main:] ~ Query requires iterating over all vertices [(__patch.id =
JAVA_PATCH_0000_006)]. For better performance, use indexes
(StandardJanusGraphTx$3:1357)
Observation : Everytime switchover happens its recreating repetitively
every-time internal typedef's because of which switchover time is more than 1 hr
Please suggest how to go about the same ,
was:
We have tried setting up Apache atlas(2.2.0) in High availability mode in
Kubernetes with separate pods for two instance.
When we simulate a failure on active instance, the failover from passive
instance from active instance is taking over an hour to complete the
transition. This happens when we have created around 300 entities whereas when
we simulated the same failover on fresh install of Atlas, it completed the
transition in 15 minutes.
This is the log before the failover for the passive instance:
The logs after the failover is in the attached file. This is the failover
during a fresh install. It is doing the similar process but taking more time
when there are more entities created in Atlas.
Observation : Everytime switchover happens its recreating repetitively
every-time internal typedef's because of which switchover time is more than 1 hr
Please suggest how to go about the same ,
> Everytime changeover happens ,Atlas is taking almost an hour to switch over
> from PASSIVE to ACTIVE in HA mode.
> --------------------------------------------------------------------------------------------------------------
>
> Key: ATLAS-4742
> URL: https://issues.apache.org/jira/browse/ATLAS-4742
> Project: Atlas
> Issue Type: Bug
> Reporter: Jagadesh Kiran N
> Priority: Major
> Attachments: application.log, image001.png
>
>
> We have tried setting up Apache atlas(2.2.0) in High availability mode in
> Kubernetes with separate pods for two instance.
>
> When we simulate a failure on active instance, the failover from passive
> instance from active instance is taking over an hour to complete the
> transition. This happens when we have created around 300 entities whereas
> when we simulated the same failover on fresh install of Atlas, it completed
> the transition in 15 minutes.
>
> This is the log before the failover for the passive instance:
> The logs after the failover is in the attached file. This is the failover
> during a fresh install. It is doing the similar process but taking more time
> when there are more entities created in Atlas.
> Warning : [main:] ~ Query requires iterating over all vertices [(__patch.id =
> JAVA_PATCH_0000_006)]. For better performance, use indexes
> (StandardJanusGraphTx$3:1357)
>
> Observation : Everytime switchover happens its recreating repetitively
> every-time internal typedef's because of which switchover time is more than 1
> hr
> Please suggest how to go about the same ,
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
