[Gluster-users] Replica bricks fungible?

Zenon Panoussis oracle at provocation.net
Sun Jun 13 15:53:29 UTC 2021


> Based on your output it seems that add-brick (with force) 
> did not destroy the already existing data, right ?

Correct, it did not.

> Have you checked the data integrity after the add-brick ? 

I only checked cursorily with 'ls thisdir' and 'ls thatdir'
to see if things looked OK and they did. But it can't hurt to
also try checksumming some (non-)random files, so I did that
too. They all match.

Note: I did the checksums on /gfsroot/volname/.glusterfs/hex/hex/uuid
and not on the fuse mounts, partly for speed and partly to
eliminate the risk of the client reading from a non-local
brick:

count=0; for f in `find /vol/gfs/gv0/.glusterfs/[0-9a-f][0-9a-f]/ -type f`; do count=$(($count + 1)); echo $count; if [ $(($count % 10000)) -eq 0 ]; then sha256sum $f >>sha`hostname`; fi; done

BTW, I saw an error in the documentation of my previous mail.
Where it says "rsync -vva /datasource/blah 127.0.0.1:gv0/" it
should say "rsync -vva /datasource/blah /mnt/fuse/gv0mountpoint/".
Sorry for that.



More information about the Gluster-users mailing list