[
https://issues.apache.org/jira/browse/GEODE-3780?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16949831#comment-16949831
]
ASF subversion and git services commented on GEODE-3780:
--------------------------------------------------------
Commit b70dcebef7ab4703058546b67dc3c395fc5d3b39 in geode's branch
refs/heads/develop from Bruce Schuchardt
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=b70dceb ]
GEODE-3780: Fixing monitoring of suspected member after passed final check
- Suspected member is never watched again after passing final check
- This restores our original behavior (pre-1.0) behavior of performing a
final check on a member if UDP communications with that member fail.
- We now also send exonoration messages to all other members if a suspect is
cleared. We need to do that because another member may have sent a
Suspect message that was ignored because the suspect was already
undergoing a final check.
- I also noticed that our tcp/ip final check loop was performing more than
one check in many cases because the nanosecond clock has a coarse
granularity. A socket so-timeout based on the millisecond clock was
timing out but the nanosecond clock didn't line up with that timeout and
caused the "for" loop to make another attempt. I changed that loop to
convert the nanosecond clock value to milliseconds.
> suspected member is never watched again after passing final check
> -----------------------------------------------------------------
>
> Key: GEODE-3780
> URL: https://issues.apache.org/jira/browse/GEODE-3780
> Project: Geode
> Issue Type: Bug
> Components: membership
> Reporter: Bruce J Schuchardt
> Assignee: Bruce J Schuchardt
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.11.0
>
> Time Spent: 3h
> Remaining Estimate: 0h
>
> In a network-down test we saw a node on the losing side of the network
> partition perform final checks on members on the winning side. One of the
> final checks mysteriously succeeded
> [info 2017/09/17 12:24:45.552 PDT
> gemfire1_rs-FullRegression-2017-09-15-21-00-35-client-10_8941 <Geode Failure
> Detection thread 4> tid=0x128] Final check failed but detected recent message
> traffic for suspect member
> 10.32.109.252(gemfire3_rs-FullRegression-2017-09-15-21-00-35-client-16_6135:6135)<v2>:1026
> [info 2017/09/17 12:24:45.552 PDT
> gemfire1_rs-FullRegression-2017-09-15-21-00-35-client-10_8941 <Geode Failure
> Detection thread 4> tid=0x128] Final check passed for suspect member
> 10.32.109.252(gemfire3_rs-FullRegression-2017-09-15-21-00-35-client-16_6135:6135)<v2>:1026
> After this the suspected member was never checked again and the losing side
> failed to shut down.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
