MLDonkey Forum Index
Homepage •  Bugs •  Tasks •  Patches •  SF.net Project Page •  ChangeLog •  German forum •  Links •  Wiki •  Downloads
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 
 
Minimizing disk access when idle
Goto page Previous  1, 2
 
Post new topic   Reply to topic    MLDonkey Forum Index -> Problems with MLDonkey Client (in general, and ed2k, kad, overnet)
View previous topic :: View next topic  
Author Message
draga
neophyte


Joined: 13 Oct 2010
Posts: 14
PostPosted: Mon Nov 08, 2010 10:45 pm    Post subject: Reply with quote
Thanks. I will try with the verbosity options suggested by spiralvoice and leave it on all the night, to see what will happen.
Then, if nothing useful comes out, will try to start trying with the grep.
Will report the results soon.
Back to top
View user's profile Send private message
draga
neophyte


Joined: 13 Oct 2010
Posts: 14
PostPosted: Tue Nov 09, 2010 8:48 am    Post subject: Reply with quote
spiralvoice wrote:
draga wrote:
is there a way to ask a more verbose log than "verb"

Suggestion: "verb file share md4 swarming"


Here's a piece of the result of a overnight run:

Code:
Nov 09 07:44:27 localhost mlnet: [Ux32] flush all
Nov 09 07:44:57 localhost mlnet: [Ux32] flush all
Nov 09 07:45:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:45:27 localhost mlnet: [Ux32] flush all
Nov 09 07:45:57 localhost mlnet: [Ux32] flush all
Nov 09 07:46:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:46:27 localhost mlnet: [Ux32] flush all
Nov 09 07:46:57 localhost mlnet: [Ux32] flush all
Nov 09 07:47:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:47:27 localhost mlnet: [Ux32] flush all
Nov 09 07:47:57 localhost mlnet: [Ux32] flush all
Nov 09 07:48:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:48:27 localhost mlnet: [Ux32] flush all
Nov 09 07:48:57 localhost mlnet: [Ux32] flush all
Nov 09 07:49:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:49:27 localhost mlnet: [Ux32] flush all
Nov 09 07:49:57 localhost mlnet: [Ux32] flush all
Nov 09 07:50:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:50:27 localhost mlnet: [Ux32] flush all
Nov 09 07:50:57 localhost mlnet: [Ux32] flush all
Nov 09 07:51:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:51:27 localhost mlnet: [Ux32] flush all
Nov 09 07:51:57 localhost mlnet: [Ux32] flush all
Nov 09 07:52:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:52:27 localhost mlnet: [Ux32] flush all
Nov 09 07:52:57 localhost mlnet: [Ux32] flush all
Nov 09 07:53:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:53:27 localhost mlnet: [Ux32] flush all
Nov 09 07:53:57 localhost mlnet: [Ux32] flush all
Nov 09 07:54:07 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:54:27 localhost mlnet: [Ux32] flush all
Nov  9 07:54:53 localhost /netbsd: sd0: pack is stopped, restarting...
Nov 09 07:54:58 localhost mlnet: [Ux32] flush all
Nov 09 07:55:08 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:55:28 localhost mlnet: [Ux32] flush all
Nov 09 07:55:58 localhost mlnet: [Ux32] flush all
Nov 09 07:56:08 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:56:28 localhost mlnet: [Ux32] flush all
Nov 09 07:56:58 localhost mlnet: [Ux32] flush all
Nov 09 07:57:08 localhost mlnet: [EDK] scanned shared files, 0 files found
Nov 09 07:57:28 localhost mlnet: [Ux32] flush all


So, as far as I can see, every minute there's a scanning of shared files and a flush, 30 seconds later.
But at 07:54:53, the disk spinned up again. Nothing was different, anyway. So it doesn't seem that mlnet has tried any other disk operation. But, actually, when mlnet is killed that doesn't happen Sad
Back to top
View user's profile Send private message
spiralvoice
Sage


Joined: 06 Jan 2003
Posts: 3982
Location: Germany
PostPosted: Tue Nov 09, 2010 8:58 pm    Post subject: Reply with quote
Try increasing buffer_writes_delay to, for example, 90.
_________________
Link overview and precompiled cores here: http://mldonkey.sourceforge.net/DownloadLinks
Back to top
View user's profile Send private message
draga
neophyte


Joined: 13 Oct 2010
Posts: 14
PostPosted: Wed Nov 10, 2010 12:15 pm    Post subject: Reply with quote
ygrek wrote:
Ehm, use grep Smile With strace you can filter specific calls, i.e. strace -e open will show only open() calls. NetBSD trace should have something similar, otherwise simply grep the output for 'open', there shouldn't be many.


Compiling strace on NetBSD. Ktrace output was quite unclear Smile
Back to top
View user's profile Send private message
draga
neophyte


Joined: 13 Oct 2010
Posts: 14
PostPosted: Wed Nov 10, 2010 12:15 pm    Post subject: Reply with quote
spiralvoice wrote:
Try increasing buffer_writes_delay to, for example, 90.


Tried it. No difference at all.
Now compiling strace, will have a deep look at it.
Back to top
View user's profile Send private message
draga
neophyte


Joined: 13 Oct 2010
Posts: 14
PostPosted: Thu Nov 11, 2010 9:17 am    Post subject: Reply with quote
Performed some more tests.
Put the process under "ktruss", grepping out all the "open" operations. It works, but nothing came out: the disk spinned up without any "open" operation by mlnet. The funny thing is that it was off for 6/8 hours, then started to wake up every 10 minutes for 1 minute or so, then stopped for 2 hours, woke up 2 times in 30 mins, then stopped for 4 hours then woke up 5 times in 20 mins.
No open operations registered by ktruss. Killed mlnet, disk was off at all the time.
Don't know where to look at anymore Smile
Back to top
View user's profile Send private message
ygrek
professional


Joined: 20 Mar 2010
Posts: 521
PostPosted: Thu Nov 11, 2010 10:46 am    Post subject: Reply with quote
I don't know how netbsd manages memory (VM subsystem), but a guess - maybe it is system trying to page out mlnet process? Do you have swap? Can you make the trace(truss) and grep out obviously uninteresting syscalls (like gettimeofday or poll) and show the part around the time when disk spins up (I understand that it will be huge)? Also try running `vmstat 1` (netbsd analogue) in parallel to watch system activity. In order to analyze such whole-system affecting strange events dtrace is invaluable but needs some learning to use.
_________________
Download | Report bugs | git mirror
Back to top
View user's profile Send private message Visit poster's website
draga
neophyte


Joined: 13 Oct 2010
Posts: 14
PostPosted: Thu Nov 11, 2010 5:41 pm    Post subject: Reply with quote
ygrek wrote:
I don't know how netbsd manages memory (VM subsystem), but a guess - maybe it is system trying to page out mlnet process?


Exactly what I was thinking about. Even if swap usage was 0, i tried to disable completely the swap space. But no difference at all Sad
[/quote]

Quote:
the time when disk spins up (I understand that it will be huge)? Also try running `vmstat 1` (netbsd analogue) in parallel to watch system activity. In order to analyze such whole-system affecting strange events dtrace is invaluable but needs some learning to use.

In the weekend, I will have a deep look also this way. It will be difficult, but now I want to understand deeply what goes on Smile
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    MLDonkey Forum Index -> Problems with MLDonkey Client (in general, and ed2k, kad, overnet) All times are GMT
Goto page Previous  1, 2
Page 2 of 2

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © phpBB Group
Sourceforge.net Logo