Am 06.03.2013 10:31, schrieb Ed Greshko:
On 03/06/13 17:20, Reindl Harald wrote:
>
> Am 06.03.2013 10:03, schrieb Ed Greshko:
>> On 03/06/13 16:52, Gabriel VLASIU wrote:
>>> On Wed, 6 Mar 2013, Ed Greshko wrote:
>>>
>>>> So you feel the -u parameter (access time) isn't sufficient to keep
from
>>>> deleting directories in use?
>>> If the service did not access the private tmp directory in X day, yes.
>>>
>>>
https://bugzilla.redhat.com/show_bug.cgi?id=801943
>>>
>> I see nothing in that bugzilla which states that a service has failed due to a
directory in /var/tmp/systemd-private-* being deleted by the cron job while the
file/directory was in use.
> and THAT it what happens!
>
> if nothing is using the PrivateTmp which reflects /tmp
> for the service all is fine and you would not need it
> at all, if the service is using it you are f**ed
Shouldn't the -umc flag of tmpwatch prevent directories in use from being deleted?
define in use
it is not in use all the time until the service creates
a temp-file and if this is a long time after service
start it fails - not more and not less
If you know of a service which doesn't access its related
/var/tmp/systemd-private-* in 30 days and then subsequently fails when deleted by the
tmpwatch cron job then please bugzilla it.
tmpwatch is broken by design