Y[OpenAFS-devel] OpenAFS release team weekly meeting

[Harald Barth]

>> I don't have any partitios available but I can give up
>> swap and format that partition to ext4 and give a try.

> Removing all swap space may cause other "interesting" issues. If you
> have sufficient space left on some filesystem, you could instead
> create a file with a fileystem on it and loop mount that:

Well, with the amounts of RAM today, most workloads don't need swap.
If your process exceeds your RAM, there is often something wrong with
it and it's better it terminates right away instead of first slowing
down the whole machine to a crawl. PDC's compute nodes run without
swap and we have seen no problems from that. If you have a firefox
with 600 tabs (I recently read an error report about that) of course
your mileage may differ ;-) ;-)

> dd if=/dev/zero of=/where/space/left/image bs=1M count=2048
> mkfs.ext[234] /where/space/left/image
> mount -oloop /where/space/left/image /mnt/openafs_cache 

Yes, that's of course a good way to test as well. 

> Let's face it, systemd is upon us, and it's not that bad. What is bad is the way it's used (configured) in current distros.

Well, I don't know if I should compain about systemd or to the folks
who wrote the rule how systemd should start openafs. Result was that
systemd restarted the openafs service endlessly until there were so
many mounts on /afs that RAM was exhausted (oops ;-)

Harald.