Re: [Duplicity-talk] Fwd: GPG taking many ressources

[Henrik Olsen]

On Aug 14, 2010, at 9:32 PM, Kenneth Loafman wrote:

> Henrik Olsen wrote:
>> On Aug 12, 2010, at 2:02 PM, address@hidden wrote:
>> 
>>> On 12.08.2010 13:21, Kenneth Loafman wrote:
>>>> Jacob Godserv wrote:
>>>>> On Tue, Aug 10, 2010 at 17:25, Henrik
>>>>> Olsen<address@hidden>  wrote:
>>>>>> As I'm still having the issue, I was wondering if GPG 1.4.10
>>>>>> could be the issue, and I should try 2.x instead?
>>>> There are still some interface changes that need to be made for
>>>> 2.x, but it might work.
>>> I recall having used duplicity together with gpg2 flawlessly.
>>> Worked for me.
>> 
>> Trying with gpg2 now. Took a bit to get it to compile with the many
>> dependencies, but now duplicity can use it, as long as I change
>> self.call from gpg to gpg2 in GnuPGInterface.py.
>> 
>> But... Unfortunately I still see a heavy ressource usage, exactly
>> like with gpg. Now with gpg2 as well, and as I noticed yesterday with
>> gpg, also even if only one process is running:
>> 
>> ps aux | grep gpg root     98278  95.1  0.0  2436668    804 s000  RN+
>> 8:42PM   2:28.35 gpg2 --status-fd 5 --passphrase-fd 10 --logger-fd 3
>> --batch --no-tty --no-secmem-warning --decrypt holsen   98287   0.8
>> 0.0  2435036    528 s002  S+    8:44PM   0:00.00 grep gpg
>> 
>> On non-OSX usage, do anyone ever see gpg usage like this?
> 
> Did gpg2 solve the issue of multiple gpg processes running?  From your
> example, it looks like it did.  If so, that was the issue that I was
> trying to solve.  Every gpg process that is still open is tying up 3
> file descriptors and a very small amount of memory, but no CPU.  With
> long chains of incrementals some people were running out of open files.

Duplicity hasn't run that long yet with gpg2, but until now I see only on gpg2 
process. But I saw earlier with gpg that it could extend to more processes 
after a while, so I'll wait with any conclusions right now.

I'm just still wondering if the heavy gpg (with both gpg and gpg2) is really to 
be expected...(?) My homedir backup which I started around 50 mins ago has 
already used 36 mins of gpg2 cpu time, using the first 14 mins for copying 
signatures to local cache. After the signature copying, the log has not 
reported anything else (even with -v9) while gpg2 is crunching away. What is it 
doing, and can it really take that long? Perhaps yes, if it's recalculating 
checksums of some kind for every file in my homedir.

/Henrik