[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Query about failure of Debian 6 64 bit to swap properly



Am Mittwoch, 29. August 2012 schrieb Bret Busby:
> Hello.

Hello Bret,

> In the ongoing saga of the inability of the 64 bit version of Debian 6
> to swap properly, so that an i3 CPU with 8GB of RAM and a 40GB swap
> partition, runs about as fast as an 8086 trying to run MS Windows 3, a
> possible cause has occurred to me.
[…]
> So, my query is this; is the inability of 64 bit Debian 6, to swap
> properly, instead using increasing amounts of RAM until it runs out of
> RAM, then crashing, while having 40GB of unused swap partition
> allocated and "swappiness" set to 70, due to the inability of the file
> manager to cope with filesize greater than 1GB?

1) Debian is not unable to swap. Period. Neither Lenny, nor Squeeze, nor 
Wheezy, nor Sid.

2) So your issue must be something else. As already pointed out later in 
thread it seems to be insane amount of memory usage.

3) 40 GB of swap with 8 GB of RAM is an quite insane setup. A harddisk 
takes in about 40-100 MiB per second on sequential access. Swap accesses 
can be quite random. So its at least 10 seconds per GiB of swap or 80 
seconds per 8 GiB of swap. Usually *lots* more. And writing can be even 
slower. The machine is likely to start to crawl to a halt by anything near 
to 16 GiB of swap usage, likely even before depending on storage speed. 
Thats physics.

4) Any filemanager I have yet seen in Debian is able to copy with files 
bigger than 1 GiB. FAT32 cannot take files larger than 4 GiB with default 
cluster sizes. Some applications may still have problems with really huge 
files. (See Large File Support in linux kernel and large / huge file support 
in Ext4.)

5) You say you are inable to handle files bigger than 1 GiB. What *exactly* 
happens? What are the error message.

Please stay just by the facts.

Preliminary conclusions can be completely of track.


In a situation of starting memory pressure I ask you to:

- free -m
- or better cat /proc/meminfo
- vmstat 1 and at least 30-40 lines of it


Additionally please install and then do:

- Start atop
- Press m
- Look which processes have the highest RGROW and SWAPSZ values

It seems to my atop sorts by resident set usage which seems saner to me 
than sorting by virtual memory usage.


Also look for anything that atop marks by turquoise or especial red color. 
In in doubt try to crap some text snapshots of it. Should be quite easy 
since by default it has a 10 second delay between updates.

Ciao,
-- 
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7


Reply to: