When doing heavy disk IO on my system (16GB RAM, 6 cores, RAID0 PCIe SSD), the system stalls and eventually crashes. Normal day-to-day disk IO is fine, just prolonged read/write such as a database re-index are problematic.

Some investigation has revealed dmesg output such as:

[...] ata2: exception Emask 0x10 SAct 0x0 SErr 0x90202 action 0xe frozen
[...] ata1: SError: { RecovComm Persist PHYRdyChg 10B8B }
[...] ata1: hard resetting link
[...] ata4: irq_stat 0x00400000, PHY RDY changed

And tons of other logs for various sata devices attached to the system. When doing heavy disk IO, any other input stalls the system (moving the mouse etc) for a 5-30 seconds.

Given the errors are on multiple devices and the system seems sketchy when giving inputs from other devices/causing non-disk interrupts ... im going to assume motherboard. Temps are fine, memcheck passes. The crash is a total soft lockup rather than a kernel crash or switch-off.

Does this seem reasonable? Anything else I can do to check?

Edit: Fedora20, home server on 870A-UD3 board, Intel 910 pcie SSD + OCZ vertex + Intel X25

Edit2: Workstation is in a home office doing officy stuff like churning data in Postgresql. All very boring nothing to warrant another close vote.

migrated from serverfault.com Mar 28 '14 at 17:16

This question came from our site for system and network administrators.

  • 6
    What motherboard, what server type, what make/model of PCIe SSD, what OS/Distribution/Version, what's your sign (Capricorn), A/S/L? – ewwhite Mar 27 '14 at 21:14
  • 4
    @ewwhite message heard. Taurus, 28/M/your dreams ;) – Aiden Bell Mar 27 '14 at 21:19
  • What else have you got plugged into this system? Name every piece of hardware. – Michael Hampton Mar 28 '14 at 1:24
  • I had this kind of problems with a dying power supply. Can you try another one? – Tobias Mädel Mar 28 '14 at 10:41
  • Can try another PSU, will do this this evening – Aiden Bell Mar 28 '14 at 17:41

This is likely a system board, cabling or power supply issue. If your SATA drives, the OCZ and Intel X25, are connected to the motherboard, they're probably the culprits. There's no indication that the Intel 910 is impacted.

  • swapped the cables on the OCZ and Intel with no effect. I don't recall error logs on the 910, but the postgres tablespace is on that device. – Aiden Bell Mar 28 '14 at 17:47
  • In totally fresh machine, same lockup issue with no ata logs. Thinking this might be the 910 causing the issue. – Aiden Bell Mar 29 '14 at 12:26

Bad sata cable. I've seen this error a few times. https://bbs.archlinux.org/viewtopic.php?id=129401

  • Giving errors on multiple drives? Also, have swapped cable with no change in errors. – Aiden Bell Mar 27 '14 at 21:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.