Skip to contentSkip to conversation

br: memory squeeze!

Bridge TC Hough 3 participants 3 messages Apr 21, 2004 – Apr 26, 2004 Ethernet bridge

lore
Share & tools
lore

TC Hough

author
#0 lore raw

I set up a bridge for my network and I'm having some trouble. At first everything worked great, but after about 3 days of continuous bridging, my bridge machine died. Approximately once a second, the message: "br: memory squeeze!" would appear on the console. Keyboard input was mostly frozen, as well. I could switch virtual consoles, but I couldn't type. No traffic would pass through the bridge. When I rebooted, everything seemed fine.

I don't have a whole lot of memory on the bridge machine, but I feel like it should be sufficient, since bridging is all it ever does and I have plenty of swap space. The system log doesn't should anything I thought to be abnormal. The memory squeeze messages were not logged. Here's the output of /proc/meminfo:

total: used: free: shared: buffers: cached:

Mem:  29655040 28917760   737280        0 13750272  6123520
Swap: 67104768   434176 66670592
MemTotal:        28960 kB
MemFree:           720 kB
MemShared:           0 kB
Buffers:         13428 kB
Cached:           5700 kB
SwapCached:        280 kB
Active:           4344 kB
Inactive:        18036 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:        28960 kB
LowFree:           720 kB
SwapTotal:       65532 kB
SwapFree:        65108 kB

Any suggestions?

TC

#1 lore raw
Quoted 7 lines · On Wed, 21 Apr 2004 12:00:07 -0500 TC Hough <[email protected]> wrote:

I set up a bridge for my network and I'm having some trouble. At first everything worked great, but after about 3 days of continuous bridging, my bridge machine died. Approximately once a second, the message: "br:
memory squeeze!" would appear on the console. Keyboard input was mostly frozen, as well. I could switch virtual consoles, but I couldn't type. No traffic would pass through the bridge. When I rebooted, everything seemed fine.

You have a memory leak in the kernel. What kernel and distro are you using. Also, what kind of NIC's and other interfaces (tun, ppp, etc).

Also, what is the state of the bridge? br showstp br0
and the forwarding table? br showmacs br0

Quoted 32 lines

I don't have a whole lot of memory on the bridge machine, but I feel like it should be sufficient, since bridging is all it ever does and I have plenty of swap space. The system log doesn't should anything I thought to be abnormal. The memory squeeze messages were not logged. Here's the output of /proc/meminfo:

total: used: free: shared: buffers: cached:

Mem:  29655040 28917760   737280        0 13750272  6123520
Swap: 67104768   434176 66670592
MemTotal:        28960 kB
MemFree:           720 kB
MemShared:           0 kB
Buffers:         13428 kB
Cached:           5700 kB
SwapCached:        280 kB
Active:           4344 kB
Inactive:        18036 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:        28960 kB
LowFree:           720 kB
SwapTotal:       65532 kB
SwapFree:        65108 kB

Any suggestions?

TC

_______________________________________________
Bridge mailing list
[email protected]
http://lists.osdl.org/mailman/listinfo/bridge

#2 lore raw
Quoted 8 lines · Stephen Hemminger
Quoted 1 line · On Wed, 21 Apr 2004 12:00:07 -0500 TC Hough <[email protected]> wrote:

I set up a bridge for my network and I'm having some trouble. At first everything worked great, but after about 3 days of continuous bridging, my bridge machine died. Approximately once a second, the message: "br: memory squeeze!" would appear on the console. Keyboard input was mostly frozen, as well. I could switch virtual consoles, but I couldn't type. No traffic would pass through the bridge. When I rebooted, everything seemed fine.

Quoted 5 lines

You have a memory leak in the kernel. What kernel and distro are you using. Also, what kind of NIC's and other interfaces (tun, ppp, etc).

Also, what is the state of the bridge? br showstp br0
and the forwarding table? br showmacs br0

I agree it is probably a memory leak, but it is *possible* to get this error if you just run out of available memory. I have been able to reproduce this type of error by setting my send/receive packet socket buffers very high and then flooding 200 connections with traffic, for example. (I had a 256MB RAM machine)

You might want to make sure that your netdev-backlog settings are not too big, for example.

Thanks,
Ben

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com