Project

General

Profile

Bug #3246

ZFS I/O deadman thread

Added by Christopher Siden almost 7 years ago. Updated almost 7 years ago.

Status:
Closed
Priority:
Normal
Category:
zfs - Zettabyte File System
Start date:
2012-10-01
Due date:
% Done:

100%

Estimated time:
Difficulty:
Medium
Tags:
needs-triage

Description

Certain hardware/drivers misbehave in such a way that I/O issued to them hangs for arbitrary amounts of time (ranging from hours to forever). Since ZFS ends up waiting for these I/Os to finish before completing a transaction group the entire pool (and any process trying to do I/O on that pool) ends up hanging. In this situation it would actually be preferable to panic the system since service would return after a reboot and leave us with a crash dump that can be used to diagnose the underlying problem (the alternative being that the system continues to hang until a user notices and reboots it manually, which means we probably don't even know why the system was hung).

History

#1

Updated by Christopher Siden almost 7 years ago

  • Status changed from In Progress to Closed
  • % Done changed from 90 to 100
commit 283b84606b6fc326692c03273de1774e8c122f9a
Author: George.Wilson <george.wilson@delphix.com>
Date:   Mon Oct 29 12:08:09 2012 -0500

    3246 ZFS I/O deadman thread
    Reviewed by: Matt Ahrens <matthew.ahrens@delphix.com>
    Reviewed by: Eric Schrock <eric.schrock@delphix.com>
    Reviewed by: Christopher Siden <chris.siden@delphix.com>
    Approved by: Garrett D'Amore <garrett@damore.org>

Also available in: Atom PDF