I have seen this 3 times now - if unRAID crashes, let's say runs out of memory and kernel decides to kill the SHSF process (syslog attached), my windows XP box hangs. Never seen an XP hang that badly. Bringing unRAID back up again does not help.

 

I initiated a shutdown remotely (I run radmin server/viewer) and first off it spent 10 minutes before getting to the blue shutdown screen, now it has been hanging on the "closing all network connections..." screen for at least 35 minutes. I decided to try and wait it out this time, but must say I am losing patience now...

 

That kind of sucks!  Any ideas what can be done to XP to avoid this kind of hang (other than making sure unRAID does not crash) ?

syslog_2010-03-22.txt

Mar 23 01:59:20 FileServer kernel: 69498 total pagecache pages

Mar 23 01:59:20 FileServer kernel: 0 pages in swap cache

Mar 23 01:59:20 FileServer kernel: Swap cache stats: add 0, delete 0, find 0/0

Mar 23 01:59:20 FileServer kernel: Free swap  = 0kB

Mar 23 01:59:20 FileServer kernel: Total swap = 0kB

Mar 23 01:59:20 FileServer kernel: 130864 pages RAM - Does your system only have 128MB RAM?

Mar 23 01:59:20 FileServer kernel: 0 pages HighMem

Mar 23 01:59:20 FileServer kernel: 2299 pages reserved

Mar 23 01:59:20 FileServer kernel: 10080 pages shared

Mar 23 01:59:20 FileServer kernel: 123171 pages non-shared

Mar 23 01:59:20 FileServer kernel: Out of memory: kill process 10655 (python) score 3487 or a child - Seems your unraid server ran out of memory, which your hinting anyways.

 

If you RAM is low, I put get at lease 512MB, but RAM being cheap, get at lease 1GB RAM instead. The fact that your unraid is crashing shouldn't bring your XP machine down (for a long period of time anyways). Do you have a file on your server that your XP machine is constantly referring to, or something like a dependency or something important like that? Perhaps your trying to save a file to your server, causing it to crash?

Does your system only have 128MB RAM?

He probably has four times as much, but not enough for his addons...

He's running SABnzbd or something, and clearly running out of memory.

His XP was at that time probably hacking at the stuff that SABnzbd downloaded.

 

@MortenSchmidt:  One possible tweak comes to mind.  Telnet to your server and type this:

sysctl -w vm.min_free_kbytes=8192

... then see if you can crash your server.  If you can't, then you can make it permanent by adding it to your 'go' script.

 

Thanks for looking at my log - Yep, Sabnzbd makes my 512MB server run out of memory. I'm looking into that - btw, what is that sysctl command supposed to do  Would like to know before messing with that stuff.

To address the sabnzbd problem I have first off disabled non-essential add-ons (e-mail notify, PHP for smarthistory, unmenu and kept only the apcupsd and powerdown addons. Also, I found that filling up the queue in sabnzbd eats up lots of memory (memory consumption doubled when adding 21 nzb files on top of 3 existing ones in the queue).

 

Later this summer I will upgrade my desktop and reuse it's 2GB and core2 duo 2GHz for unRAID. Despite getting 2GB RAM, I still might want to look into replacing sabnzbd with getnzb, though (comments welcome!).

 

But back to the windows box, this is what I mainly seek advice on. You will notice from the log that also the SHSF process gets killed. From my other computer I could see the disk1, disk2, etc shares were still accessible but the user shares went down. My HTPC with windows XP was playing a 720p h264 mkv file at the time (was watching heroes), using Corecodec 1.9.5 and Mediaportal 1.0.2. So nothing fancy, just watching a file on the fileserver.

 

Any ideas for tweaks which will eliminate this windows defect, I mean it shouldn't hang this way just because a fileserver goes down. It's a big risk for the WAF factor, one thing is for the fileserver to go down (let's be honest, it can happen once in a blue moon), but the HTPC too... not good, let me tell 'ya!

 

what is that sysctl command supposed to do ?

Nothing dangerous. It will just rise a little bit the threshold below which your lazy garbage collector will start working.

I had a similar problem in the past, and I had 1GB RAM.  It stopped running out of memory after I rised that treshold.

 

what is that sysctl command supposed to do ?

Nothing dangerous. It will just rise a little bit the threshold below which your lazy garbage collector will start working.

I had a similar problem in the past, and I had 1GB RAM.  It stopped running out of memory after I rised that treshold.

 

Thanks, Will give that a try. Have a feeling I could probably still crash it if I wanted to, but if it won't hurt anything else I might as well use it.

Have a feeling I could probably still crash it if I wanted to, but if it won't hurt anything else I might as well use it.

Don't just 'use' it, see if you can crash the server after you apply it! 

That will give us a definite answer whether it solves your problem or not.

 

I can still crash the thing, it's just a matter of filling up the sabnzbd queue. It's not the sheer number of files, it's the size of them. Queuing up 6 HD movies where the nzb files are somewhat larger is enough to crash it. This is with the sysctl command you listed, and running without e-mail notify, the PHP package and unmenu.

 

Back to the main topic - has anyone got an idea why windows would hang FOREVER when unRAID crashes while watching a show residing on unraid