On 26/01/16 19:58, 'Igor Galić' via s3ql wrote:

split brain situations aren't theoretical. completely avoiding them
is sometimes impossible. http://queue.acm.org/detail.cfm?id=2655736
Sometimes you lose a machine because it's literally on fire, or just
because its hard disk(s) breaks…

Have you thought about setting --metadata-upload-interval to something lower than the default 86,400? Maybe reduce it to 3,600 which means you won's lose more than an hour's data.

Even just a shutdown of the vm (which, in theory, unmounts all file systems)
will make it so that it's impossible to mount the fs on another machine:

Are you sure that the shutdown is (a) calling umount.s3ql and (b) waiting for it to finish? It can take a very long time to return, depending on the size of the dirty cache, the size of the metadata and the speed of the line.

Regards
Cliff.

--
Cliff Stanford
Office: +44 20 0222 1666      UK Mobile:  +44 7973 616 666
Spain:  +34  952 587 666
http://www.may.be/

--
You received this message because you are subscribed to the Google Groups 
"s3ql" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to