[GLLUG] Odd SATA Issue - slightly OT

Ken Smith kens at kensnet.org
Tue Jan 8 19:33:26 UTC 2019


>>>
>>>
>>> Alain D D Williams via GLLUG wrote:
>>>> On Mon, Dec 24, 2018 at 08:51:09PM +0000, Greater London Linux UUG 
>>>> wrote:
>>>>> Hi Everyone,
>>>>>
>>>>> I have an older centos 6 system that has 3 sata disks in soft raid5.
>>>>> I've noticed that the disk performance has become very poor.
>>>>>
>>>>> I did several "dd if=/dev/sdX of=/dev/null count=10000" tests on the
>>>>> disks and found one of them performing around 1 MB/s while the
>>>>> others are at 150 MB/s or more.
>>>> A cheat fix attempt is to replace the power supply. These tend to 
>>>> fade over time
>>>> and can cause disks to play up - however I would have expected to 
>>>> see errors in
>>>> /var/log/messages
>>>>
>>>
>>> Thanks Alain, I hadn't thought of that. I'll measure some voltages 
>>> first and see what I get.
>>>
>>> :-) Ken
>>>
>>>
>>
>>
>>
> PSU voltages looked good but swapped it and the Motherboard (I have a 
> spare) & sata cable but sadly no change. Have added a spare disk and 
> done some mdadm juggling to introduce the new disk and fail the 
> suspect one. Once the raid had rebuilt the performance is back to normal.
>
> /boot & Grub loader are on the suspect disk. Have already copied /boot 
> to the new disk but need to reinstall the boot loader on that disk and 
> update fstab with the new /boot.
>
> Strange that a WD NAS disk can work fine except 100x slower than 
> normal. iostat -x was showing huge values for await on that WD device.
>
> Oh well - chalk that down to experience.
>
> :-) Ken
>
To wrap this story up - the disk has been replaced and all has returned 
to normal. Just before I replaced it one pending sector error was 
logged.  Interesting how the performance can take a dive long before 
errors are reported.

:-) Ken


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.




More information about the GLLUG mailing list