Minimizing disk access when idle

[draga]

Hi everybody. I've been using MLDonkey for years and it's my favourite p2p app.
Now I got a Alix board and installed NetBSD on it. I connected an external hard drive and it works perfectly.
The drive, after some minutes of inactivity, goes to sleep so it spares energy and noise.
When Transmission is working (without files), the drive goes to sleep and never goes up again till i put a new file on queue.
MLDonkey, instead, wakes up the disk quite often, even if there are no files. I've modified some options:
ED2K-remove_old_servers_delay to 86400.
save_options_delay to 86400
share_scan_interval to 0
I also linked the log file to /dev/null

It goes better, but i see it still wakes up the disk. Sometimes it does it after 6/7 hours, sometimes many times per hour.
Have you any hints or suggestions? The idea would be of keeping mlnet running at all the times, knowing it won't wake up the disk if no files are shared or downloading.

[spiralvoice]

Have a look at option backup_options_delay and whether via option
interval in web_infos files are re-downloaded regularly.
_________________
Link overview and precompiled cores here: http://mldonkey.sourceforge.net/DownloadLinks

[draga]

[ygrek]

http://mldonkey.sourceforge.net/Web_infos
_________________
Download | Report bugs | git mirror

[draga]

Thank you. Checked. Some of them (like GeoIP) are updated just at core start, others once every some days.

[Balamutick]

Hi, draga.

I am inspired by your question (plus it was a bit of time) and wrote an article in wiki for your question. In two languages. Article in Russian original, because the article on English can be a little grammatically incorrect (who can - please correct).

Tell me please, last post helped you finally solve your problem?

P.S. That article on English
_________________
Russian community Mldonkey. Русское сообщество MLdonkey.

[draga]

[draga]

Well, I did some tests.
The disk spins up in an unpredictable way. Sometimes it stays off for hours, sometimes it wakes up every few minutes. There doesn't seem any apparent reason.
Is it possible that, not resharing at regular intervals, the disk spins up when another user asks us a chunk of a file that according to mlnet is still here but, instead, has been removed?

[spiralvoice]

[draga]

Thanks spiralvoice. Just configured to "verb" debug level and to log on syslog (on a ramdisk, of course )
Now will wait for interesting messages. Meanwhile, i have done a script that performs a "reshare" every time a file gets removed by the "incoming", in order to be sure that no missing files are still considered by mlnet.
Will post the results as soon as i will have some news.

[draga]

Mmm...disk is spinned up without apparent reason. The "verb" option also doesn't suggest anything. Look at the log here:

[ygrek]

strace mlnet and look for the open syscalls to see what files it accesses. I think bsd analogue of strace is called ktrace but I am not familiar with its options.
_________________
Download | Report bugs | git mirror

[draga]

[spiralvoice]

[ygrek]