Note: This web page was automatically created from a PalmOS "pedit32" memo.
3ware maxtor notes
3ware notes
maxtor notes
ESMF promicro notes: 3ware and maxtor
3ware 8506 card: 2t limit
3ware 9506: no 2t limit. $11k?
ESMF 3Ware cards should see good performance improvement by using a
2.6.5 or better linux kernel. We were using 2.4.x kernels.
Bunch of Maxtor 251G SATA disks. Model is MaXLine Plus II, 250GB SATA/150
HDD, 1.5G/s
Using 3Ware card models 8506-12 8506-8
In case it matters, we were using SuperMicro motherboards in the
promicro computers.
Dead disk/disk firmware/3ware firmware mentions:
1) 18 Aug 2004, Anthony of Promicro requested that I get him the version
off of one of the 3ware cards
5) August 27, Dan Stromberg resent the firmware revisions to ProMicro
2) Mentioned in teleconference notes for 2004-09-01
3) Another dead disk mention in telecon notes for 2004-09-20
6) 20 Sep 2004, Dan Stromberg discovered that 3ware driver and
3ware firmware were downrev due to initrd 3ware driver hiding the
/lib/modules 3ware driver
7) 20 Sep 2004: Dan confirms that upgrading the 3ware driver also
upgraded the 3ware firmware
8) 22 Sep 2004: Anthony confirms 3ware updates "ESMFSN01, 02 and 03 have
now been updated with the latest 7.7.1 firmware and drivers"
4) Dead disk, and a dead RAID volume, mentioned in telecon notes for 2004-09-27
9) 5 Nov 2004: Anthony indicates that Maxtor wasn't forthright about
need to upgrade Maxtor drives' firmware. Anthony announces plan to
bring a PC to UCI, to do Maxtor firmware upgrades with
11) 2004: 11/8 or 11/9: Suggested (by Anthony of Promicro) dates for
Maxtor firmware upgrades
12) 11/09/2004: Dan writes to Ross Aiken of IBM: Rajiv and Promicro and
I are all hoping that upgrading the firmware on the hard disks is going
to eliminate some or all of our errors
13) Nov 9, 2004: Vince Balsamo of IBM writes: Anthony should be onsite
at UCI tomorrow to repair systems and update firmware
14) 10 Nov 2004: Ross writes "Each of the 3ware adapters in all systems
were checked and updated as required to bring up the latest scsi adapter
firmware levels required."
15) 11/10/2004: Francisco writes "Anthony Shipp is/was on-site upgrading
the firmware on a subset of the Maxtor drives that are having problems.
Per Maxtor's
recommendation, Anthony is not upgrading the firmware on all Maxtor
drives. I am slightly concerned that some drives with the older
firmware will begin having problems over time and need to be upgraded
at a later date, but I am willing to defer to Maxtor's recommendation
at this time."
10) 11/12/2004: Ross Aiken of IBM confirms that all 3ware cards have had
their firmware upgraded
16) 2005-11-12: Anthony of Promicro: sn01: no degraded arrays. sn02's
SCSI-3 port 4 was bad, and was replaced with a new disk. SCSI-0, port
9 and 11, had old code of firmware, I flashed them. SN03: 2 disks on
SCSI-3 that had old firmware code, I flashed those. SN04: I did nothing
to this machine. SN05: N/A. SN06: On SCSI-0, one disk dead, despite
good firmware, on SCSI-1 a disk had old firmware but refused to be
flashed - disk replaced.
17) 19 Nov 2004: Dan writes: We've seen "SCSI errors" from time to time.
Upgrading the firmware on our Maxtor drives has hopefully gone at least
part way to resolving this.
18) 23 Nov 2004: Dan writes: looks like sn01's SCSI ID 3, Array Unit 0,
port 1 is offline. Are you up for upgrading the firmware on it, if you
haven't already?
3Ware, by way of Promicro, provided us with some system tweaks that
should optimize I/O with the 3Ware cards. However, it is my belief that
these tweaks impacted the buffer cache behavior, and because Lustre
doesn't use the usual linux buffer cache, the tweaks likely didn't help.
The 3Ware card's firmware can be updated by compiling firmware upgrade
support into the linux driver for the 3Ware card. However, the firmware
on the Maxtor SATA disks could not (at the time, this functionality is
planned for later) be upgraded through the 3Ware RAID cards.
Info on Maxtor SATA disk firmware releases, given to UCI by Promicro:
"code" on disk describes firmware rev
bw0 all to hw0
ew0 can't update to h, so not updated
hw0 good
w0 good
can also see rev in 3dmd
Anthony says e's are sometimes troublesome