[ https://issues.apache.org/jira/browse/CASSANDRA-10864?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15066688#comment-15066688 ]
Jim Witschey commented on CASSANDRA-10864: ------------------------------------------ Here: https://www.mail-archive.com/user%40cassandra.apache.org/msg44644.html you mentioned that you were going to look at commitlog behavior. Did you get any new information from that? And were you able to watch {{ttop}} at all, as described here: https://www.mail-archive.com/user%40cassandra.apache.org/msg44589.html [~yukim] This seems to be related to, or at least effecting, repair. Could you have a look at the email thread and see if you can tell what the issue is? > Dropped mutations high until cluster restart > -------------------------------------------- > > Key: CASSANDRA-10864 > URL: https://issues.apache.org/jira/browse/CASSANDRA-10864 > Project: Cassandra > Issue Type: Bug > Reporter: Jeff Ferland > > Originally raised and investigated in > https://www.mail-archive.com/user@cassandra.apache.org/msg44586.html > Cause is still unclear, but a rolling restart has on two occasions since been > performed to cope with dropped mutations and timed-out reads. > Pattern is indicative of some kind of code quality issue possibly involving > locking operations. Stack flame graphs do not show a clear difference between > restarts. -- This message was sent by Atlassian JIRA (v6.3.4#6332)