-
Type: Bug
-
Resolution: Done
-
Priority: Major - P3
-
Affects Version/s: 1.5.1
-
Component/s: None
-
None
After performing an rs.stepdown() and then an rs.reconfig() to set the previous master node to hidden/priority=0, traffic did not properly move to the new master automatically, I had to restart the app servers.
This worked fine under 1.4.x.
Mongostat looked like this for a while:
*0 *0 51 *0 0 1|0 0 538g 1080g 37.8g 0 0.1 0 0|0 1|0 13k 495b 216 set1 SEC 05:27:14
And on the new master
0 51 0 0 0 2 0 348g 700g 37.5g 0 0 0 0|0 2|1 16k 542k 203 set1 M 05:26:14
After an app server restart, it looks more normal:
*10 1 1|54 *0 43 2|0 0 538g 1080g 37.8g 2 12.7 0 0|0 1|0 2k 20k 174 set1 SEC 05:32:15
9 20 48 0 62 1 0 348g 700g 37.5g 1 5.3 0 0|0 2|1 23k 237k 233 set1 M 05:32:39
Just tried it on another replica set, similar behavior. No ruby exceptions were thrown, but no data was written until the app servers were manually restarted.
- related to
-
RUBY-381 ReplSetConnection does not correctly discover new master after a disconnect
- Closed