Couchbase Server

autofailover 2 minutes delay when failover timeout is set to 30 seconds

Details

Description

when one of the nodes go missing and stop responding autofailover will take an extra minute or two to automatically failover the node

more information from Aliaksey about this bug:


Just tested normal shutdown case. Indeed it was longer than expected. Our implementation needs extra 2 hearbeat periods to actually confirm that other nodes are up so it explains 10 seconds more then 30. But in my testing it was around 50 seconds when timeout was set to 30 seconds.

Issue Links

depends on

Bug - A problem which impairs or prevents the functions of the product. MB-4733 autofailover 2 minutes delay when failover timeout is set to 30 seconds

  • Blocker - Blocks development and/or testing work, production could not run.
  • Resolved - A resolution has been taken, and it is awaiting verification by reporter. From here issues are either reopened, or are closed.

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
Steve Yen added a comment -
Last update on this was awhile ago. Any news/status on this?
Show
Steve Yen added a comment - Last update on this was awhile ago. Any news/status on this?
Hide
Permalink
Steve Yen added a comment - - edited
Alk's fixing this for 2.0 homogeneous clusters for now
Show
Steve Yen added a comment - - edited Alk's fixing this for 2.0 homogeneous clusters for now
Hide
Permalink
Steve Yen added a comment -
from dipti... at the moment 30 secs is the minimum timeout for auto failover timeout and that satisfies customers
Show
Steve Yen added a comment - from dipti... at the moment 30 secs is the minimum timeout for auto failover timeout and that satisfies customers
Hide
Permalink
Farshid Ghods added a comment -
Is this merged/fixed ?
Show
Farshid Ghods added a comment - Is this merged/fixed ?
Hide
Permalink
Aleksey Kondratenko added a comment -
1.8 and 2.0 should work fine w.r.t. this particular problem
Show
Aleksey Kondratenko added a comment - 1.8 and 2.0 should work fine w.r.t. this particular problem
Hide
Permalink
Aleksey Kondratenko added a comment -
correction: I was referring to branch-18. I.e. 1.8.2 which has quick failover
Show
Aleksey Kondratenko added a comment - correction: I was referring to branch-18. I.e. 1.8.2 which has quick failover
Hide
Permalink
Perry Krug added a comment -
Is that in the upcoming 1.8.1?
Show
Perry Krug added a comment - Is that in the upcoming 1.8.1?
Hide
Permalink
Aleksey Kondratenko added a comment -
No. 1.8.1 was (at least initially) supposed for simple fixes. Running replicators on replicas was deemed not simple. We were not allowed to have this on 1.8.1. 1.8.2 if it will be released will have this fix.
Show
Aleksey Kondratenko added a comment - No. 1.8.1 was (at least initially) supposed for simple fixes. Running replicators on replicas was deemed not simple. We were not allowed to have this on 1.8.1. 1.8.2 if it will be released will have this fix.
Hide
Permalink
Peter Wansch added a comment -
Let's move this into the 2.0 branch. I assume the 1.8.2 fix has been reviewed and tested already. Thanks.
Show
Peter Wansch added a comment - Let's move this into the 2.0 branch. I assume the 1.8.2 fix has been reviewed and tested already. Thanks.
Hide
Permalink
Aleksey Kondratenko added a comment -
2.0 automagically has this as everything in branch-18 was merged to 2.0
Show
Aleksey Kondratenko added a comment - 2.0 automagically has this as everything in branch-18 was merged to 2.0

People

  • Assignee:
    Aleksey Kondratenko
    Reporter:
    Farshid Ghods
Vote (0)
Watch (3)

Dates

  • Created:
    Updated:
    Resolved: