You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@accumulo.apache.org by GitBox <gi...@apache.org> on 2022/12/21 13:25:36 UTC
[GitHub] [accumulo] ivakegg opened a new issue, #1169: balancing strategy
ivakegg opened a new issue, #1169:
URL: https://github.com/apache/accumulo/issues/1169
The balancer currently will assign out the tablets given their initial locations per the "last" cf in the metadata, and the will slowly move tablets to get a final steady state. In many deployments we are starting up a large accumulo cluster and it will take an hour or more to attain a steady state. There needs to be a mechanism by which we delay the initial assignment (most likely handled by #1167) and then to determine a final steady state before doing the initial assignments. This will help in the case where we are expanding the cluster considerably and bringing it up for the first time.
--
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.
To unsubscribe, e-mail: notifications-unsubscribe@accumulo.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [accumulo] ivakegg commented on issue #1169: balancing strategy
Posted by GitBox <gi...@apache.org>.
ivakegg commented on issue #1169:
URL: https://github.com/apache/accumulo/issues/1169#issuecomment-1361471659
I see that when a tserver unloads a tablet, it clears the current location so that it is no longer available upon startup. I guess that needs to stay in place otherwise many things would break. looking for another option here.
--
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.
To unsubscribe, e-mail: notifications-unsubscribe@accumulo.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [accumulo] ivakegg commented on issue #1169: balancing strategy
Posted by GitBox <gi...@apache.org>.
ivakegg commented on issue #1169:
URL: https://github.com/apache/accumulo/issues/1169#issuecomment-1361318474
Instead of updating the last location, I am going to add an option to allow accumulo restarts to be based on the "loc" entry instead of the "last" entry. I am hoping that coupled with the delayed balancing will help to significantly improve our startup times.
--
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.
To unsubscribe, e-mail: notifications-unsubscribe@accumulo.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [accumulo] ivakegg commented on issue #1169: balancing strategy
Posted by GitBox <gi...@apache.org>.
ivakegg commented on issue #1169:
URL: https://github.com/apache/accumulo/issues/1169#issuecomment-1361552774
Adding a property to preserve the current location in the last location when the tservers are unloaded/suspended.
--
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.
To unsubscribe, e-mail: notifications-unsubscribe@accumulo.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [accumulo] ctubbsii commented on issue #1169: balancing strategy
Posted by "ctubbsii (via GitHub)" <gi...@apache.org>.
ctubbsii commented on issue #1169:
URL: https://github.com/apache/accumulo/issues/1169#issuecomment-1405284059
This is addressed by #3142
--
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.
To unsubscribe, e-mail: notifications-unsubscribe@accumulo.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [accumulo] asfgit closed issue #1169: balancing strategy
Posted by "asfgit (via GitHub)" <gi...@apache.org>.
asfgit closed issue #1169: balancing strategy
URL: https://github.com/apache/accumulo/issues/1169
--
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.
To unsubscribe, e-mail: notifications-unsubscribe@accumulo.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [accumulo] ivakegg commented on issue #1169: balancing strategy
Posted by GitBox <gi...@apache.org>.
ivakegg commented on issue #1169:
URL: https://github.com/apache/accumulo/issues/1169#issuecomment-1361310452
So, the only thing that was actually done to resolve the issue with accumulo restarts with LOTs of tablets (e.g. million+) was the delay of assignments pending so many tservers being available (#1158). That helps, but does not solve the issue. We still have at least an hour or two before balanacing settles down. This is primarily because the "last" location does not represent a balanced state in our systems. This significantly impacts operations.
--
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.
To unsubscribe, e-mail: notifications-unsubscribe@accumulo.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org