Red Hat Bugzilla – Bug 905203
glusterfs 3.3.1 volume heal data info <something> not accurate.
Last modified: 2013-03-22 01:50:21 EDT
Description of problem:
output from checking heal status doesn't seem to be accurate.
Version-Release number of selected component (if applicable):
Tried test just once.
Steps to Reproduce:
1. Setup 3 node (1 brick each) replicate cluster with the following settings:
2. Start rsync from client to filesystem.
3. Bring node 1 down for 5-10 minutes.
4. Bring node 1 back up
5. Bring node 2 down for 5-10 minutes.
6. Bring node 2 back up
7. Bring node 3 down for 5-10 minutes.
8. Bring node 3 back up.
9. Start & stop rsync a few times as a user might when testing/changing options.
Files seem to heal but "gluster volume heal newdata info heal-failed" and "split-brain" return a bunch of results.
Trying an "rsync -avcnP /mnt/gluster/bricks/1/home/ root@server:/mnt/gluster/bricks/1/home" (dry run,via checksum) returns nothing to be done so it looks like the data is actually synched.
I'd expect after the heal has finished to show nothing in heal-failed or split-brain.
This is my first-try using a quorum of 2 out of 3 nodes. It seems taht the heal-failed/split-brain simply pulls info from logs but doesn't show what the current state is. Is this expected?
*** This bug has been marked as a duplicate of bug 864963 ***