i think unless iet has a memory leak, it should not come to oom so
easily.
* consider it is a SYNC write. which means no dirty data after each
write and thus page cache can be reclaimed always
* consider it only support 1 connection per session, so if there is only
1 session, then it should be ok to allocated at most 128K page buffer to
hold a write, and then write to disk, and then free the page.
so i think there are something else here.
can u check the memory usage from the beginning of u run bonnie and see
what happen in slab and memory usage?
i can pass mem=64MB on my box and see what happen. :) but not now. i am
under deep pressure now.
ming
On Sun, 2005-07-24 at 23:48 +0900, Harald Kubota wrote:
> Hi,
>
> I got my hands on a small 'server' not constisting of much more than USB
> ports, an FastEthernet port, a SH4 CPU, 64MB of memory
> and a CF boot medium. It's originally supposed to be a Samba server for
> USB disks. And it does that job quite nicely.
> See http://www.iodata.jp/prod/storage/hdd/2004/usl-5p/index.htm
>
> But this is hardly interresting, so the reason why I actually bought it,
> is to use it as a small iSCSI target.
> However I quickly run into a serious problem: ietd's memory usage.
>
> I can export a USB disk fine via ietd. I can connect to it. I can do
> stuff like formatting, copying small files (and not so small files too).
> But "bonnie++ -s 2000" kills it with those messages:
>
> Hi,
>
> I got my hands on a small 'server' not constisting of much more than USB
> ports, an FastEthernet port, a SH4 CPU, 64MB of memory
> and a CF boot medium. It's originally supposed to be a Samba server for
> USB disks. And it does that job quite nicely.
> See http://www.iodata.jp/prod/storage/hdd/2004/usl-5p/index.htm
>
> But this is hardly interresting, so the reason why I actually bought it,
> is to use it as a small iSCSI target.
> However I quickly run into a serious problem: ietd's memory usage.
>
> I can export a USB disk fine via ietd. I can connect to it. I can do
> stuff like formatting, copying small files (and not so small files too).
> But "bonnie++ -s 2000" kills it with those messages:
>
> http://www.iodata.jp/prod/storage/hdd/2004/usl-5p/index.htm
> Free pages: 1068kB (0kB HighMem)
> Active:19 inactive:451 dirty:0 writeback:421 unstable:0 free:267
> slab:627 mapped
> :0 pagetables:6
> DMA free:1068kB min:1024kB low:1280kB high:1536kB active:76kB
> inactive:1804kB pr
> esent:65536kB pages_scanned:1056 all_unreclaimable? no
> lowmem_reserve[]: 0 0 0
> Normal free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB
> present:0kB pag
> es_scanned:0 all_unreclaimable? no
> lowmem_reserve[]: 0 0 0
> HighMem free:0kB min:128kB low:160kB high:192kB active:0kB inactive:0kB
> present:
> 0kB pages_scanned:0 all_unreclaimable? no
> lowmem_reserve[]: 0 0 0
> DMA: 3*4kB 6*8kB 5*16kB 1*32kB 0*64kB 1*128kB 1*256kB 1*512kB 0*1024kB
> 0*2048kB
> 0*4096kB = 1068kB
> Normal: empty
> HighMem: empty
> Swap cache: add 689, delete 689, find 15/26, race 0+0
> Free swap = 248856kB
> Total swap = 248968kB
> Out of Memory: Killed process 741 (btnctrl).
> oom-killer: gfp_mask=0xd0
> DMA per-cpu:
> cpu 0 hot: low 14, high 42, batch 7
> cpu 0 cold: low 0, high 14, batch 7
> Normal per-cpu: empty
> HighMem per-cpu: empty
>
> Many of them. It's clear to me, this is caused by the machine running
> low on memory. 64MB is not the world, but I wonder
> if there's no less drastic way to handle low memory situations. Even
> when I had 1GB of RAM, I might run out of memory
> if enough and enough fast traffic happens, right? However, even when
> pushing the dual P3 box with 1GB RAM I have
> running ietd, I never got an oom situation. Just luck or is it
> impossible to get in that case?
>
> Directly after booting and starting ietd, 18MB memory are used, leaving
> about 40MB for buffers. Is that not enough?
> How much is enough?
>
> If I can enable debugging on initiator/target side if useful.
>
> I realize I am not waking in high-performance-land with this machine,
> but it's quick enough to saturate its FastEthernet port.
>
> Thanks,
> Harald
>
>
>
> -------------------------------------------------------
> SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
> from IBM. Find simple to follow Roadmaps, straightforward articles,
> informative Webcasts and more! Get everything you need to get up to
> speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
> _______________________________________________
> Iscsitarget-devel mailing list
> Iscsitarget-devel@xxxxxxxxxxxxxxxxxxxxx
> https://lists.sourceforge.net/lists/listinfo/iscsitarget-devel
-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
|
