| 1 |
>>>Nothing. The logs simply stop when the machine does and are no longer written |
| 2 |
>>>even though the kernel is operable enough to respond to pings and initiate |
| 3 |
>>>TCP sessions. |
| 4 |
>>> |
| 5 |
>>> |
| 6 |
>My experience is that this is related to running out of memory, and have |
| 7 |
>replicated it several times on boxes with limited memory. To test I had |
| 8 |
>two ssh sessions open running `watch uptime` and `watch free`, and then |
| 9 |
>hammered the box -- the load shot up and free memory was rapidly |
| 10 |
>shrinking toward 0 and then the box froze. |
| 11 |
> |
| 12 |
> |
| 13 |
> |
| 14 |
I would have to agree, check out memory usage. |
| 15 |
I had the same exact thing happening to a server of mine. |
| 16 |
The cause was the server running out of memory. |
| 17 |
I had accidentally removed the swap partition out of fstab... |
| 18 |
|
| 19 |
|
| 20 |
The only other time I had something like that happen was when one of my |
| 21 |
co-workers was pinging the complete 172.16.0.0/12 network. |
| 22 |
) ...including the broadcast addresses. :^( |
| 23 |
Most of the machines had dual nics in the same physical network, and |
| 24 |
they were getting a huge arp request flood. |
| 25 |
The arp tables filled up and that was that, no log entries or nothing. |
| 26 |
The machines would seem to crash. |
| 27 |
I believe that the current kernels handle that better now, but hey who |
| 28 |
knows for sure. |
| 29 |
|
| 30 |
|
| 31 |
You mentioned that your switch is a ProCurve. |
| 32 |
Which model is it? We use the 5300 series, and they log problems on the |
| 33 |
ports when the switch notices them. |
| 34 |
Check out the switch logs, sometimes it can really help. |
| 35 |
|
| 36 |
Regards, |
| 37 |
Scott |
| 38 |
|
| 39 |
-- |
| 40 |
gentoo-server@g.o mailing list |
Archives