• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 2408
  • Last Modified:

PowerEdge 1800 SATA Cerc 6 squeals with high-pitch tone during format

I have a PE1800 (Cerc 6) that I bought with 3 * SATA 750GB disks but recently upgraded to 1TB * 6 SATA disks.
The card cannot make a >2TB array, so there are  two arrays: Cerc Arr012 and Cerc Arr345, each made of 3 * 1TB disks in RAID 5 for a total of 1.9 TB per array.

I installed and configured this server as an R-Snapshot backup server, but during initial testing, I realized that I only had 2TB of space to use for backups because LVM could only go up to 2TB in a volume. I needed to repartition, so I backed up the system using rsync and am trying to reinstall RHEL5.

At first, I tried to put the LVM-usable disk space from Arr012 and Arr345 into the same Volume Group so that I could have one big /srv partition for backups:
Arr012: sda1-/boot + sda2-swap + sda3-LogVol00
Arr345: sdb1-LogVol00
LogVol00 = swap + /(root) + /var + /tmp + /srv
However, when Disk Druid got to the formatting stage, a continuous high-pitched beep or whine or alarm sound emanated from the case. It just kept going for about 10 minutes while I was Googling the issue. Finally, I shut the machine down and started the installation over again.

I then tried this:
Arr012: sda1-/boot + sda2-LogVol00
Arr345: sdb1-LogVol01
LogVol00 = swap + /(boot) + /var + /tmp + /srv
LogVol01 = /srv2

I divided it up like the above and then set it to format. It did not squeal like the first try, but it does squeal for a little and then go quiet as it formats. And the format is taking forever. I went home after 20:00, and it was still formatting /srv2. I cannot tell whether it is stalled or not, but I would liek to know why the high-toned squeal.

Should I worry about this sound?
Can I rely on this server in this condition?
Should I give up on this machine?
0
shinmaikeru
Asked:
shinmaikeru
2 Solutions
 
BigSchmuhCommented:
The high-pitched beep may be due to a read error that made the drive enters in a too long error recovery process which forced the CERC card to disconnect it...and beep !

More details below...

Large SATA drives are suffering of TWO major features:

1/ TLER (or CCTL) are missing feature on Desktop editions
Desktop drive editions delay their answer when a read error occurs to retry and retry the read ... too long for the hw raid card not to consider the drive is dead/not responding anymore (it's back at reboot time...but it still has some unreadable sectors somewhere waiting for you !)

2/ "Unrecoverable bit read error per bits read" ratio is too high for large drives
Most of the Desktop edition have a 1 per 10 exp 15 ratio (sometimes 1 per 10 exp 14 for the worst SATA Desktop drives).
This means, you have a 0.8% probability to get an unrecoverable read error per 1TB read (8% for those drives with a 1 per 10 exp 14 ratio)

NB: Some SAS drives reach a 1 per 10 exp 16 "Nonrecoverable Read Errors per Bits Read" ratio for their 2,5" models
   http://www.seagate.com/docs/pdf/datasheet/disc/ds_savvio_15k.pdf
==> This is an "Enterprise" usage of course


Conclusion:
-No RAID 5 with SATA drives while they keep having such a high failure rate !
-Doing RAID 6 with 4 drives is way slower than RAID 10 and cost the same
-Doing RAID 6 with 5-6 drives is not a bad idea if you align the partition and try to use array block size equals to your client io block size
0
 
shinmaikeruAuthor Commented:
Thank you for the information. Just to clarify, these are enterprise-grade drives that I carefully chose and ordered directly from a Seagate dealer, and they are 3.5", but I appreciate the extra information. In researching this, I also learned that the SATA Cerc 1.5/6 sold in the  Dell PE1800 has a terrible reputation.

However, the problem turned out to be a failed drive in the second array, Arr345. Number 4 was bad. I was formatting for a new installation when it happened, so there was no RAID monitoring software working. When I rebooted, I entered the RAID configuration utility and it clearly said that HDD04 was bad.

I am now working on getting it replaced and buying a spare. My initial plan was to have a hot spare, but the 2TB limit that necessitates making two R5 arrays prevents me from using the hot spare, so I will have to have a backup sitting in the server room.

Thank you again.
0

Featured Post

Industry Leaders: We Want Your Opinion!

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

Tackle projects and never again get stuck behind a technical roadblock.
Join Now