Whoops, I actually ran the command verbatim and then tried it without
the trailing part in case the package had been updated and I might be
able to get it via a "meta package". Basically a shot in the dark that
didn't work. That cut/paste was from the last thing I tried...

I switched back over to the sil3124 based cards and was able to get it
working pretty solid. The way I can "re-create" the issue is by creating
a RAID 5 set with the 3 1Tb WD drives and copying data from a 500gb
external drive over to it. A few hundred megs into the copy, the console
starts spewing the above messages and after a few mins the transfer
stalls when XFS deactivates the filesystem (rightfully so it seems). I
can actually run a bonnie++ and it won't trigger the bug, but the big
copy of files will do it every time. It could take longer because the
1Tb drives are actually 5400 RPM and somewhat slow.

I pulled the 3114 cards and replaced them with the two 3124 cards.  I
hooked the three WD drives up and MD recognized them by UUID on boot. It
actually did and XFS repair and mounted the volume back up after the
boot, but I didnt' trust it. I did a "dd if=/dev/zero of=/dev/md0
bs=1024k count=100" to wipe the start of the md block device and did a
"mkfs -t xfs /dev/md0" to re-build the filesystem. I was able to copy
the entire 480 gig from the external over to the new 1.9Tb volume. I was
also seeing errors relating to the swap partition that happens to sit on
the 250 gig seagate drive also off the 3114 controllers.

It seems pretty stable now. I am going to blast it (zero out all three
disks) and re-create the RAID volume (an overnight affair I am afraid),
re-create the filesystem and give it another try. If its stable, then I
am going to stick to the 3124 based controllers. The only problem I have
with them is that they are 2 internal interfaces and 2 external (eSATA)
interfaces and I have 10 internal SATA hotswap slots in this case....
They have solder pads for two more internal SATA connectors and they use
0 ohm SMD jumper resistors. If I am stuck using these cards I may get
frisky and try de-soldering the SATA jacks from the 3114 cards, moving
them to the 3124  cards and moving the SMD jumper resistors so I can
have 4 internal ports on each:).  Yes, I can solder SMD, but its a
bit*h....

I am willing to blow up my setup one more time for the sake of science
and the greater good to see the newer kernle can get the 3114 cards
working. I am also willing to send the 3114 cards off to someone who
works on this driver to get it stable. It looks like it will happen with
different types of drives and I am soon to be quite attached to my 1.9Tb
of storage so I will not send those out.

Those 3114 cards are quite pervasive (and cheap) and it would be nice to
have them work reliably. Personally, unless something changes I am going
to focus on the 3124 cards at this point for my budget SATA needs.

-- 
filesystem corruption on xfs using sata_sil (sil3114)
https://bugs.launchpad.net/bugs/156612
You received this bug notification because you are a member of Ubuntu
Bugs, which is the bug contact for Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to