3,342 reputation
614
bio website geekosaur.dreamwidth.org
location Akron, OH
age 49
visits member for 3 years, 1 months
seen Nov 11 '13 at 18:39

Twitter: geekosaur

(back online although still looking...)

I'm an old-time Unix/Usenet geek; you can get a good idea of my checkered past by searching for my name in Google Groups. (Include the first name, though, or you'll get my cousin Russ (http://www.eyrie.org/~rra)....)

Presently I am unemployed and looking at jobs in the Toronto area. Most recently I was a senior system administrator for Carnegie Mellon University's Department of Electrical and Computer Engineering, a department of the Carnegie Institute of Technology half of CMU. Most of my work is infrastructure — the stuff most people don't notice unless it's broken. My job is to make sure people continue to not notice it.

I have been suffering from gradually worsening sinus problems for the past 10 years, and I was recently diagnosed with dysphoric bipolar syndrome. A fair amount of this journal is about the latter. More recent items are tagged so they can be avoided. I may at some point go back and fix the earlier ones.

In the 90s (that is, before the dysphoric bipolar got bad enough that I had trouble focusing in projects) I was involved with a number of open source projects. I'm starting to find my way back into the open source world now that I'm regaining reliable control of my faculties.


Mar
31
answered SSH Fails, /bin/bash: No such file or directory
Mar
30
comment Linux: Force different set of directory/file permissions
No, I mean that a directory created in a g+s directory will get the correct group but will not itself be made g+s, so files created within that directory won't inherit the group. POSIX ACLs deal with this by providing masks of permissions and ownership that will be inherited by files and directories, rendering the setgid thing unnecessary; but some Linux distributions (I don't know about RHEL/CentOS specifically) don't support POSIX ACLs out of the box.
Mar
30
answered Linux: Force different set of directory/file permissions
Mar
25
comment Ubuntu load average spikes but CPUs are idle
You're welcome. (I noted NFS and iSCSI from the loaded kernel modules in your kernel error message, btw.)
Mar
25
comment Ubuntu load average spikes but CPUs are idle
Something like sudo tcpdump -i (interface) host (nfsserver) or host (iscsitarget) or ... I can't be more specific because, well, the specifics are what we're trying to find.
Mar
25
comment Ubuntu load average spikes but CPUs are idle
Problem is, so am I, since I'm not there. :) I would prefer wireshark over tcpdump if it's installed. Set it to filter on the addresses of the NFS servers and iSCSI targets, so you're not inundated with traffic. (Tracing the session over which you're running the network trace is a nice little self-DoS. :) Beyond that, well, we're both hunting for clues. As to your provider, have they actually bothered to look at the logs or are they just blowing you off? (On second thought, you probably can't answer that.)
Mar
25
comment Ubuntu load average spikes but CPUs are idle
Okay, so at this point you've done pretty much everything you can do from your end. You could run tcpdump or wireshark to see if there are any obvious NFS or iSCSI errors or timeouts, but in that case your service provider's logs should show it too and don't require someone to be monitoring the network trace.
Mar
25
comment Ubuntu load average spikes but CPUs are idle
Sorry, I should have specified that it was log entries before the CPU lockup was detected.
Mar
25
comment Ubuntu load average spikes but CPUs are idle
If there's any other possibly related log information around the CPU lockup, it might be interesting to see; NFS and/or iSCSI might be involved, there's no way to know with the information you've provided so far but either could lead to VFS deadlocks. Also, consider syslog over the network to another machine (see syslog.conf(5)); it might make it easier to catch the problem without constant monitoring.
Mar
25
comment Ubuntu load average spikes but CPUs are idle
That stands a very good chance of being related, yes. The symptoms you're describing are consistent with a VFS deadlock, which is why I keyed onto the snapshot. Your update confirms that... well, "deadlock" is the least of the problems. :) And the stack trace strongly suggests that either you're hitting a Xen bug or Xen is tripping over something else failing (which would hopefully be in your provider's logs).
Mar
25
comment Ubuntu load average spikes but CPUs are idle
See edited response for details. You need to find a way to make the snapshot more lightweight.
Mar
25
revised Ubuntu load average spikes but CPUs are idle
add more detail
Mar
25
answered Ubuntu load average spikes but CPUs are idle
Mar
25
answered What tool do you recommend to track changes on a Linux/Unix server
Mar
24
answered _www password on Mac OS X
Mar
23
answered gmail/email backup server - how'd i go grabbing emails off imap, and serving them out again?
Mar
21
comment Xvnc4 started from xinetd only displays empty gray X screen
If you never had a reason to run gdmsetup (for example, to set up autologin), it probably won't exist; just create it with the [xdmcp] stuff in it.
Mar
20
comment Xvnc4 started from xinetd only displays empty gray X screen
For that, you're almost right; there's no listener for -broadcast, though, as gdm doesn't enable XDMCP by default. You need to modify /etc/gdm/custom.conf to enable it. This Arch Wiki page describes the change to make; restarting gdm may be different on your distribution.
Mar
20
answered Xvnc4 started from xinetd only displays empty gray X screen
Mar
19
answered wireshark not seeing all packets on WiFi