Le mer. 19 déc. 2018 à 12:22, Alexis BRENON <brenon.alexis@xxxxxxxxx> a écrit : > > Le mer. 19 déc. 2018 à 11:25, Wols Lists <antlists@xxxxxxxxxxxxxxx> a écrit : > > > > On 19/12/18 08:13, Alexis BRENON wrote: > > >> OMG - I was just looking for your version of mdadm, and I noticed - > > >> > Seagate Baracudas! Can you let us know (a) what version of mdadm, and > > >> > (b) have you fixed the timeout problem! If you haven't done (b) then > > >> > that's probably your problem! > > >> > > > > I do not have access to my computer right now but if I remember well, > > > my system is a Debian stretch and I upgraded it recently. So mdadm > > > should be in version 3.4-4. I will confirm that as soon as possible. > > > > Okay. So do as the wiki says and upgrade. 3.4 has a bunch of known bugs, > > although they're probably not responsible this. > > Okay, I will compile v4.1 this evening. No particular risk of > incompatibilities with other packages on my debian stretch ? > I upgraded mdadm without any problem. > > > > > > I "fixed" the timeout problem yesterday, while reading the wiki saying > > > "If you didn't fix timeout, do it right now!" ^^ My 3 SG Baracudas > > > reported a disabled SCT Error Recovery, and my WesternDigital reports > > > an unsupported command. > > > > > I used the script provided in the wiki to set the timeout to 7s on the > > > Baracudas and to 180ms for my WD but the latter is not preserve > > > accross reboot as reported in smartctl sum-up (see below). > > > > > Okay, the Barracudas don't support ERC - I think you can't enable it, so > > setting it 7s probably did nothing. Set the kernel timeout to 180 for > > the Barracudas also. > > > > Setting to 7s return 0 as output code and smartctl now reports > SCT Error Recovery Control: > Read: 70 (7,0 seconds) > Write: 70 (7,0 seconds) > for all my Barracudas drives. Is smartctl kidding me ? Running the Brad Campbell's timeout script report all my Barracuda disks as good and all my WD as bad. > > > Cheers, > > Wol > > Kind, > Alexis Now my question are: Is it safe to assemble my array with my old drives ? Or do I need to replace my sdb (which reports a read failure and has been last updated on 2018-12-08) and/or my sde (which do not report any SMART failure, but a lower event count)? Is it safe to run my array in Read only mode for some weeks, waiting for new drives delivery (probably a 0.5TB or 1TB WD Red)? Which wiki page to follow: Replacing a failed drive, Assemble run or Recovering a damaged RAID? Kind regards, Alexis.
