-
Notifications
You must be signed in to change notification settings - Fork 3.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
HBASE-28608 More sensible client meta operation timeout default #6000
Conversation
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
…t operation timeout
fixed spotless check |
🎊 +1 overall
This message was automatically generated. |
🎊 +1 overall
This message was automatically generated. |
🎊 +1 overall
This message was automatically generated. |
🎊 +1 overall
This message was automatically generated. |
For me this is reasonable. But anyway this is a behavior change, better start a discuss thread on the dev list to see if others have other opinions. Thanks. |
Thank you for reviewing @Apache9 . I was originally under the impression that the default behavior was unintentional given the documentation in the hbase reference and the discussion in the work done in HBASE-24956 which added a new critical dependence on meta operation timeout in the locate region/meta lookup codepath in 2.x, but this may not have been unintentional as I originally thought. There seem to be two very different dependencies on meta operation timeout on 2.x:
I am not familiar with how the locate region codepath on branch 3 works. But for branch 2 and the I think it may be worth decoupling the |
On branch-3+, we have removed all the sync client code and reimplement sync client on top of async client. In async client,the locate region timeout is controlled by meta operation timeout, but since we are asynchronous, the upper layer could get a timeout even if the locating region operation has not completed yet, so there is no problem. On branch-2, it is a problem since we run all the operations in one thread. Thanks. |
Thank you for the explanation @Apache9 . I have opened HBASE-28730 to propose decoupling the meta operation timeout from the branch-2 userRegionLock/locate region codepath. If my proposal there sounds reasonable, I can work to get that implemented, and once that big branch-2 userRegionLock dependence on the property is gone, it should be simpler to reason about the implications of a new default for meta operation timeout and I can bring this PR up on the devlist for discussion. |
Following up on my last comment, I originally was thinking to block this work on HBASE-28730 , but that issue itself is blocked on HBASE-27781 (which has patch available and is pending review), but even with HBASE-28730 completed there will still exist a dependency on meta operation timeout property , so I think would be good to get this in without blocking on HBASE-28730 + HBASE-27781, I will start a devlist discussion for the proposed change here shortly and see if there is support |
Started dev list discussion here - https://lists.apache.org/thread/6q918prw2n6g90x1gfom4qf1hr3blxvv |
Hi @Apache9 , from dev list thread started one week ago the proposal has received two +1s (so three +1s in total including your approval). |
…t operation timeout (apache#6000) Co-authored-by: Daniel Roudnitsky <droudnitsky1@bloomberg.net> Signed-off-by: Duo Zhang zhangduo@apache.org
https://issues.apache.org/jira/browse/HBASE-28608