On current jogamp.org server, we have experienced silent GNU/Linux SW Raid-1 block mismatches.
[root@jogamp sysconfig]# cat /sys/block/md2/md/mismatch_cnt
The semantics of GNU/Linux SW Raid are inconclusive, i.e. it could be nothing,
but also a silent degradation (block failure).
Since the only reliable filesystem w/ proven block failure detection via sha256 checksums
including healing via redundancies (mirror or raidz) is ZFS,
we need to move to a new box w/ preferably ECC Ram and 2x Enterprise HDD for mirroring setup.
A work in progress is documented here (doc will move to git later on):
The move will probably happen within next week, i.e. June 3rd - 7th.
Machine bring-up is done:
1st rsync is performed.
- Migrate users, ..
- Adapt old configs /data to new software versions
- Move old configs / services over
Switch DNS to new IP-Address !
@All: Please refrain from writing data to jogamp.org until I give the green light tomorrow... Thank you!
(In reply to comment #2)
> @All: Please refrain from writing data to jogamp.org until I give the green
> light tomorrow... Thank you!
this includes: git, wiki, bugzilla, jenkins, user accounts .. etc! :)
write access allowed again, files moved and following services reinstated:
logging, move backup files and users, mysql, procmail, bogofilter, sasl2, dovecot, sendmail,
git, apache and jabot