Welcome! Log In Create A New Profile

Advanced

Disconcerting - very - total failure

Posted by fxm 
fxm
Disconcerting - very - total failure
June 01, 2010 06:13PM
I just sat down at my system.
I noticed that the PB icon was not in the systray.
Investigation showed that PB was not running - no surprise, but worrisome.
The surprise - and the real worry - came when I restarted PB. The .conf and .conf.bak files were both gone.
The log file was truncated (at about 8meg). I say 'truncated' because the last line in what's left was an utterly routine entry from 6 days ago (same as the timestamp on the file before the restart) with no error indication whatever. No .dmp file anywhere.

I don't exactly watch for the systray icon all the time, but it is inconceivable that PB could have been not running for 5+ days without my noticing. I'm assuming - at least partly because it makes me feel less vulnerable - that the crash occurred some time today.

So, there you have it: a major failure which left zero traces.

As I say, very disconcerting.
Anonymous User
Re: Disconcerting - very - total failure
June 01, 2010 07:00PM
Hey there fxm,

I agree mate, it is disconcerting the issue you are having, I leave my pc running all the time also, with sometimes weeks in between reboots, I have not noticed the same issue, hopefully when Markside or one of the other dev's see this they will have an answer for all of us.

Do you want to upload the files you do have, so they can have a look.

Cheers

Freelandr
fxm
Re: Disconcerting - very - total failure
June 01, 2010 07:11PM
Not really.
Nothing on my system when I restarted PB would be of any use in debugging. I checked very carefully.
The truly annoying thing was that after a very long period of solid operation by PB I recently stopped making periodic copies of the .conf file (which I used to do routinely when events like this were common [back around 181]).
Re: Disconcerting - very - total failure
June 10, 2010 08:51AM
avatar
Hmm, that's bizarre.

The code is not overwriting peerblock.conf.bak until it has successfully started (it's the last thing it does prior to logging the "Starting message-loop" message).  By this point we should have been assured that peerblock.conf was readable, and that the config we loaded from it was good.

It also looks as though we weren't at all logging to peerblock.log for some reason, which I don't understand.  There's really not a whole lot of magic going on there.

And for the .conf file to be deleted as well . . . I just don't get it.

Let's think out loud for a minute...  It's possible that we successfully loaded peerblock.conf, then were unsuccessful at getting a handle to peerblock.log, but didn't error out because of it.  We're doing any "fail-on-error" stuff there, since we figured that having PeerBlock running without logging was more desirable than refusing to let PeerBlock run at all just because we couldn't log what we were doing.  Maybe our attempt to write to peerblock.conf.bak hit a problem, such that it caused us to bomb out.  And then on the way down, we attempted to write our config out to peerblock.conf, but hit that same problem.

Then again, that would have meant that PeerBlock was not running for 5+ days, which you've said is extremely unlikely.  Given that, I really am not too sure what else could have caused this problem.  Especially since we're not touching peerblock.conf.bak at any other spot in the program.  And actually, when we're writing out our config (either to the .conf or to the .conf.bak), we're actually writing out to a .tmp file first and then overwriting the original file with that .tmp one after successfully writing to it.  And problems opening peerblock.log too?  Hmmm...

The one common thread between all these symptoms is that they're all disk related.  Is it possible that you have some latent disk-corruption somewhere?  I could see that causing these sorts of problems.  Unlikely, I know, but I'm really at a loss to explain this.

        ---  Mark  ---




Lead developer of PeerBlock
fxm
Re: Disconcerting - very - total failure
June 10, 2010 10:42AM
MarkSide Wrote:
-------------------------------------------------------
> at a loss to explain this.

Me too.

Thanks for your comments. Given the lack of evidence I couldn't expect any more.

I'm watching much more closely now, and if anything similar happens I'll try to save as many pieces as possible.

I don't think a HD failure is all that likely. I have been thinking about HIPS and similar processes. The only persistent problem I have been having with PB is the occasional bizarre failure during the list update process.
Sorry, only registered users may post in this forum.

Click here to login