ZFS: Failed pool causes system to hang

[Chris BeHanna]

On Mar 24, 2013, at 13:54 , Quartz <quartz at sneakertech.com> wrote:

>>> [...situations in which you *have* to reboot...]
> 
> [...all well and good, except for cases where someone has to physically hit a button on a remote server...]
> 
>> People who run servers remotely yet lack this capability are
>> intentionally choosing [snip]
> 
> Before you get up on a high horse and preach at me, consider a couple things:
> 
> 1) Yes I can set that up, but this is a test box on my desk right now.
> 
> 2) A hard reset is a hard reset is a hard reset. I'm not bitching that I have to physically walk over to the machine, I'm bitching that *THAT I HAVE TO RESET IT*. Being able to reset it remotely is NOT an acceptable solution or workaround, and has no bearing on my problem.

	Sure, there are bugs that inhibit the ability to reboot from command-line.  We shouldn't shrug them off, but we should also acknowledge that the software is very complicated, and rooting out such bugs takes time.  Plus, this is a volunteer project.

	That said, there are some chassis that will let you drop to lights-out management and force a reboot without having to physically press a button, provided you have a way to reach the lights-out management console.  This often means, at least, a serial line concentrator somewhere ("conserver" is simply awesome), or KVM-over-IP.

	Worst case, you have IP-addressable PDUs into which you can telnet and hard-reset a particular outlet, to power-cycle a machine that is truly, irrevocably stuck, without having to have a human walk over and stick a paper clip into the reset hole.  This should be a last resort, though, as abrupt power-cycling shortens hardware life.

-- 
Best,
Chris BeHanna
chris at behanna.org