Thank you!

So I probably I will have to use the negative etimeout version to ensure
that no estimate will take longer than a given time, regardless of how
many estimates the planner is running.

Maybe this should be corrected in future manual versions.


Thanks again,
        Ralf







Paul Bijnens schrieb:
> On 2007-09-04 16:13, Ralf Auer wrote:
>> Hi Paul,
>>
>> a 'amgetconf NewSetup etimeout' returns '900'. So, it seems to be
>> configured correctly.
>>
>> I came to the conclusion that my server still waits for the estimate by
>> issuing the 'amstatus' command. It told me, that the server is still
>> 'waiting for estimate' for one host, all other hosts were in 'estimate
>> done' state.
>> And last night I watched my backup running, and the timeout for that
>> host occured after exactly 1 hour, 15 minutes, and, be surprised, 4
>> seconds. Don't know, if theses 4 seconds are the same 4 seconds you can
>> see on your backup, I don't mind, but 1 hour and 15 minutes for sure is
>> more than 30 minutes, a.k.a. 900 sec times 2 DLEs.
>>
>> For tonight I changed the etimeout value to -3600, so that the timeout
>> should occur after one hour. Just to see if this etimeout setting works
>> at all in my setup.
> 
> Aha, did some digging in the source code, and it seems that the
> documentation is not really in sync with the code.
> 
> The "etimeout" is not multiplied by the number of DLE's for that host
> but by the number of estimates that planner will run for that host.
> 
> The config that I had my "etimeout" go off after 4*900 seconds for
> a 4 DLE host, was configured for archives, doing only full backups
> and using "strategy noinc" in the dumptype to avoid estimate runs
> for anything but full dumps as well.  So my number of estimates was
> equal to the number of DLE's.
> 
> A normal run of planner like yours above, runs more estimates (up to
> 3 per DLE).  And each estimate adds "etimeout" to the total timeout.
> So etimeout did work, but 2 DLE's, each running 2 estimates
> (probably lvl 0, and lvl N), results in 3600 seconds total timeout.
> 
> 

-- 

Ralf Auer       
Physics Institute IV                    Office: 2.137
University of Erlangen-Nuremberg        Phone:  +49-9131-8527087
Erwin-Rommel-Str. 1                     Fax:    +49-9131-15249
D-91058 Erlangen, Germany               Ralf.Auer AT physik.uni-erlangen DOT de