Chris's Wiki :: blog/tech/QuietSATADegradation Commentshttps://utcc.utoronto.ca/~cks/space/blog/tech/QuietSATADegradation?atomcommentsDWiki2010-05-26T12:42:39ZRecent comments in Chris's Wiki :: blog/tech/QuietSATADegradation.From 69.113.211.148 on /blog/tech/QuietSATADegradationtag:CSpace:blog/tech/QuietSATADegradation:fb65c59eeefd8f442bd7db384259dc57578d995eFrom 69.113.211.148<div class="wikitext"><p>When I first started dating my fiancee, her desktop computer was dragging ass like you wouldn't believe. Some of her friends had run all kinds of cleanup on it, pruning out all the traces of spyware, malware and pre-installed Dell crap and stopping just short of blowing away and reinstalling the whole OS.</p>
<p>I took a look at it, did some of the more obvious checks, and after a day or so, I finally decided to run HD Tune's benchmark on it.</p>
<p>It peaked at THREE megabytes/sec throughput.</p>
<p>Just like in your situation, there were no SMART errors, no device resets, nothing super off-kilter with the drive counters, and the only thing apparently wrong was that the disk was inexplicably running like crap but ostensibly chugging along just fine.</p>
<p>This is something that the vendors will need to begin tracking -- having it abstracted away from us by RAID controllers and SAN storage processors doesn't make the process easy for people trying to roll their own baseline monitor.</p>
<p>--Jeff</p>
</div>2010-05-26T12:42:39ZFrom 201.95.160.158 on /blog/tech/QuietSATADegradationtag:CSpace:blog/tech/QuietSATADegradation:1868e205ddb24a4c1f14c39c54032f909b7b423cFrom 201.95.160.158<div class="wikitext"><p>With a population of over 8k SATA disks, we've noticed that quite regularly. However we monitor the amount of media errors that are reported by the RAID controller and replace the disks as soon as possible. That specific threshold is not easy to calculate but we've come to the conclusion that over 100 media errors is usually bad (if they are scattered around, not localized).</p>
<p>The biggest problem is one a disk is working so badly but won't fail, so operations are retried on it for a long time. NFS clients will feel that.</p>
<p>Giovanni</p>
</div>2010-05-26T11:13:05Z