Re: There are currently problems with host "talisker"

On Mon, Dec 05, 2022 at 12:25:59AM +0000, Andy Smith wrote: What happened was, the SAS controller (which is built in to the motherboard) did something strange and stopped responding: Dec 5 00:00:20 talisker kernel: [15457308.417397] sd 0:0:1:0: attempting task abort!scmd(0x000000000eb85a6f), outstanding for 7040 ms & timeout 7000 ms Dec 5 00:00:20 talisker kernel: [15457308.417397] sd 0:0:1:0: [sdb] tag#2370 CDB: ATA command pass through(16) 85 08 2e 00 d0 00 01 00 00 00 4f 00 c2 00 b0 00 Dec 5 00:00:20 talisker kernel: [15457308.417397] scsi target0:0:1: handle(0x000a), sas_address(0x4433221101000000), phy(1) Dec 5 00:00:20 talisker kernel: [15457308.417397] scsi target0:0:1: enclosure logical id(0x500304801ce84801), slot(1) Dec 5 00:00:20 talisker kernel: [15457308.417397] scsi target0:0:1: enclosure level(0x0000), connector name( ) Dec 5 00:00:52 talisker kernel: [15457339.698559] mpt3sas_cm0: In func: mpt3sas_scsih_issue_tm Dec 5 00:00:52 talisker kernel: [15457339.699439] mpt3sas_cm0: Command Timeout Dec 5 00:00:52 talisker kernel: [15457339.700221] mf: Dec 5 00:00:52 talisker kernel: [15457339.700221] Dec 5 00:00:52 talisker kernel: [15457339.700221] 0100000a Dec 5 00:00:52 talisker kernel: [15457339.700221] 00000100 Dec 5 00:00:52 talisker kernel: [15457339.700221] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700248] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700249] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700249] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700249] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700249] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700249] Dec 5 00:00:52 talisker kernel: [15457339.700249] Dec 5 00:00:52 talisker kernel: [15457339.700249] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700249] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700258] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700258] 00000000 Dec 5 00:00:52 talisker kernel: [15457339.700258] 00000943 Dec 5 00:00:52 talisker kernel: [15457339.700258] Dec 5 00:01:02 talisker kernel: [15457349.938722] mpt3sas_cm0: sending diag reset !! Dec 5 00:01:03 talisker kernel: [15457351.223529] mpt3sas_cm0: diag reset: SUCCESS Dec 5 00:01:03 talisker kernel: [15457351.293977] mpt3sas_cm0: CurrentHostPageSize is 0: Setting default host page size to 4k Dec 5 00:01:18 talisker kernel: [15457366.578292] mpt3sas_cm0: _base_display_fwpkg_version: complete Dec 5 00:01:18 talisker kernel: [15457366.578296] mpt3sas_cm0: _base_display_fwpkg_version: timeout Dec 5 00:01:18 talisker kernel: [15457366.579119] mf: Dec 5 00:01:18 talisker kernel: [15457366.579119] Dec 5 00:01:18 talisker kernel: [15457366.579121] 12000001 Dec 5 00:01:18 talisker kernel: [15457366.579123] 00000000 Dec 5 00:01:18 talisker kernel: [15457366.579124] 00000000 Dec 5 00:01:18 talisker kernel: [15457366.579125] 00000000 Dec 5 00:01:18 talisker kernel: [15457366.579126] 00000000 Dec 5 00:01:18 talisker kernel: [15457366.579127] 00000000 Dec 5 00:01:18 talisker kernel: [15457366.579129] 00000000 Dec 5 00:01:18 talisker kernel: [15457366.579130] 00000100 Dec 5 00:01:18 talisker kernel: [15457366.579131] Dec 5 00:01:18 talisker kernel: [15457366.579131] Dec 5 00:01:18 talisker kernel: [15457366.579132] 0758a000 Dec 5 00:01:18 talisker kernel: [15457366.579134] 00000022 Dec 5 00:01:18 talisker kernel: [15457366.579135] 00000100 Dec 5 00:01:18 talisker kernel: [15457366.579136] 40000000 Dec 5 00:01:18 talisker kernel: [15457366.579137] Dec 5 00:01:18 talisker kernel: [15457366.579143] mpt3sas_cm0: mpt3sas_base_hard_reset_handler: FAILED Dec 5 00:01:18 talisker kernel: [15457366.579154] sd 0:0:1:0: task abort: FAILED scmd(0x000000000eb85a6f) Dec 5 00:01:18 talisker kernel: [15457366.579162] sd 0:0:3:0: attempting task abort!scmd(0x00000000b20a175c), outstanding for 65088 ms & timeout 30000 ms Dec 5 00:01:18 talisker kernel: [15457366.579169] sd 0:0:3:0: [sdd] tag#2380 CDB: Read(16) 88 00 00 00 00 00 a4 fc 47 50 00 00 00 08 00 00 Dec 5 00:01:18 talisker kernel: [15457366.579173] scsi target0:0:3: handle(0x000c), sas_address(0x4433221103000000), phy(3) Dec 5 00:01:18 talisker kernel: [15457366.579176] scsi target0:0:3: enclosure logical id(0x500304801ce84801), slot(3) Dec 5 00:01:18 talisker kernel: [15457366.579179] scsi target0:0:3: enclosure level(0x0000), connector name( ) Dec 5 00:01:18 talisker kernel: [15457366.579182] sd 0:0:3:0: No reference found at driver, assuming scmd(0x00000000b20a175c) might have completed …and then pages and pages more of much the same. Drives sda, sdb, sdc and sdd are SSDs that are attached to the onboard SAS ports and these hold all the customer VM storage on that server (your xvd* block devices that aren't archive storage). They had all disappeared so no I/O was possible for VMs. THe host itself still worked because its storage is on other devices. I tried for a while to reset the controller, make it re-scan etc, but all of this just met timeouts so in the end I had to send a forcible poweroff sysctl to every running VM. Some of those either didn't complete their poweroff or are set to ignore that sysctl, so for those I then had to send a "destroy" (like yanking power). There wasn't really anything that I or the VMs could do to ensure a clean shutdown as they could do no I/O at all. On boot no problems were evident. The SAS controller could see all of its drives. The RAID arrays assembled again without incident and VMs were able to start up again. I think this will have effectively been like a power loss on the storage, which your filesystems and applications should hopefully be able to handle without corruption. One customer VM is showing as down but when I had a look at the console it seemed likely to be a configuration error in the VM rather than anything related to this incident. After the host booted I made a mistake and accidentally broke IPv4 networking to talisker and its VMs for about 15 minutes. I didn't notice because I was using IPv6 myself, and I wasn't paying close attention to alerts. Once I started going through checking everything had recovered properly I did notice and fixed the issue. I don't know if the above SAS controller issue was a software bug or an indicator of hardware failure or some freak occurrence. We will keep an eye on things and if it ends up being necessary we'll move customers to a different host. If any customer on talisker is still experiencing problems, please check your Xen Shell consoles and if the solution is not then evident, please email support(a)bitfolk.com to open a support ticket. Apologies for the disruption, and for the error that lengthened the outage by about 15 minutes. Thanks, Andy -- https://bitfolk.com/ -- No-nonsense VPS hosting