Re: [OT] Solaris crashes itself when /tmp is full

From: GreyCloud (mist_at_Cumulus.com)
Date: 02/13/04 

Date: Fri, 13 Feb 2004 16:48:55 GMT

Andrew Harrison SUNUK Consultancy wrote:

> Bill Gunshannon wrote:
>
>> In article <402BA1B8.18896617@blueyonder.co.uk>,
>> Tim Llewellyn <tim.llewellyn@blueyonder.co.uk> writes:
>>
>>>
>>> Roy Omond wrote:
>>>
>>>
>>>> Any comments ?
>>>>
>>>> A flabbergasted Roy Omond
>>>> Blue Bubble Ltd.
>>>
>>>
>>> Sure, it is a unix "weakness". Some syadmins have cron jobs to
>>> monitor and clean up
>>> /tmp.
>>
>>
>>
>> Here we go again.
>>
>> No, it is not a Unix weakness. None of the numerous servers I run now
>> will crash if /tmp fills up and none that I remember in the past (which
>> goes back a couple of decades). The worst that is likely to happen is
>> that some application (vi comes immediately to mind) might not work
>> properly.
>>
>> I am curious enough about this that I plan to set up a machine with
>> Solaris 9 and test this myself. Even if it turns out to be true, it
>> is easily remedied and probably would have been, before the fact, by
>> any really experienced Unix Sysadmin.
>>
>
> Sun supplies SunMC with Solaris, one of the agents is a diskspace
> monitor, setting a threshold on /tmp and then finding and deleting
> all files over a certain age is simple to impliment.
>
> As I said earlier this seems to be a people and process problem.
>

Odd. Seeing that this is an obvious problem, why doesn't the o/s itself
warn that the limits are being approached? I noticed that when / itself
gets full, the o/s emits a FS full message and won't let any users
login. The o/s still stays up under this condition.
Can I set high water marks that the o/s can then emit a warning message?