Xavi Hernandez

@mohit84 do you know what's the major contributor to the performance improvement ? __at__ functions or the removal of the lock ?

I've done a quick check of the self-heal logs, and it seems there are no issues. Healing seems to be progressing. There have been around 8 million files healed since...

> Thanks for your help, Xavi. > > > How many files are there in the volume ? is it possible that there are more than 40 million files with...

> The hosts were taken out of production soon after the initial problem, on December 10. There should be no data being written to the volume. In that case the...

Hi @dalwise. How is it going ? Is the self-heal still progressing on the arbiter brick or have you stopped it ? if it's still running, probably it should have...

> Self-heal is ongoing. It is currently at over 64M and increasing by about ~50k per hour. I thought it would have discovered all directories by now. I was too...

@dalwise when you do configuration changes, you should also restart the self-heal processes started manually by you because they are not under the control of glusterd to apply the changes....

> Just to double check that we are on the same page: I have been running the commands you suggested and starting the extra glusterfs processes on the arbiter node,...

> Hi @xhernandez, I'm definitely glad that we clarified that and I'm sorry for not bringing it up earlier. Adding extra processes in the other two nodes caused a huge...

> I tried running `gluster volume heal vol_name full`, but that didn't seem to have any effect. This command just starts a full scan of the volume checking every file,...