| 2011-08-23 19:39:32 |
Shawn Usry |
description |
This is moreso a bug with error messaging - or at least - needs a bit of cleanup. I think I might have had a polluted environment from a previous.
Set up a 2 node environment. Configured to use the newer rsync_wan for joiner resync. For testing, I brought down the 2nd node, then back up again. Cluster correctly identified that that the joining node needed a state transfer. Transfer proceeded, but then near the end I received an error message from the SENDING side that the receiving side had "too many open files" - a typical problem - but upping the limit on BOTH sender and receiver did not resolve the issue.
It turned out the issue was due to the fact that the file "RPM_UPGRADE_HISTORY", in /var/lib/mysql, had root:root permissions, and could not be overwritten. Changing permissions to mysql:mysql resulted in the state transfer completing on the next run.
I'm not sure how the original file permissions were set. The RPM's were originally installed as root however. |
This is moreso a bug with error messaging - or at least - needs a bit of cleanup, or at least a look at permissions during state transfers.
Set up a 2 node environment. Configured to use the newer rsync_wan for joiner resync. For testing, I brought down the 2nd node, then back up again. Cluster correctly identified that that the joining node needed a state transfer. Transfer proceeded, but then near the end I received an error message from the SENDING side that the receiving side had "too many open files" - a typical problem - but upping the limit on BOTH sender and receiver did not resolve the issue.
It turned out the issue was due to the fact that the file "RPM_UPGRADE_HISTORY", in /var/lib/mysql, had root:root permissions, and could not be overwritten. Changing permissions to mysql:mysql resulted in the state transfer completing on the next run.
I'm not sure how the original file permissions were set. The RPM's were originally installed as root however. |
|
