[
https://issues.apache.org/jira/browse/MESOS-8002?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16187172#comment-16187172
]
Alex Lee commented on MESOS-8002:
---------------------------------
This is crashing our Chronos instances in Mesos cluster which was running fine
before the Mesos 1.3.0 upgrade. Leads me to believe this is more an issue with
libmesos than the individual framework.
> Marathon can't start on macOS 11.12.x with Mesos 1.3.0
> ------------------------------------------------------
>
> Key: MESOS-8002
> URL: https://issues.apache.org/jira/browse/MESOS-8002
> Project: Mesos
> Issue Type: Bug
> Components: master
> Affects Versions: 1.3.0
> Environment: macOS 10.12.x
> Reporter: Alex Lee
>
> We upgraded our Mesos cluster 1.3.0 and run into the following error when
> starting Marathon 1.4.7:
> ```
> I0823 17:19:17.498087 101744640 group.cpp:340] Group process
> (zookeeper-group(1)@127.0.0.1:57708) connected to ZooKeeper
> I0823 17:19:17.498652 101744640 group.cpp:830] Syncing group operations:
> queue size (joins, cancels, datas) = (0, 0, 0)
> I0823 17:19:17.499153 101744640 group.cpp:418] Trying to create path
> '/mesos/master' in ZooKeeper
> Assertion failed: (0), function hash, file
> /BuildRoot/Library/Caches/com.apple.xbs/Sources/cmph/cmph-6/src/hash.c, line
> 35.
> ```
> This was reported in: https://jira.mesosphere.com/browse/MARATHON-7727
> Interestingly, Marathon was able to start in the same cluster on macOS
> 10.11.6 host. We were suspecting it's OS version issue initially and open the
> issue with Apple. But macOS team responded that there may be a regression in
> mesos. The assertion is being raised in libcmph that libmeso.dylib invokes
> with providing invalid input and the hash functions in libcmph don’t look
> like they’ve changed between 10.11.6 and 10.12.6, at least with respect to
> that assert(0) being around.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
