ssh-agent stops responding
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
seahorse (Ubuntu) |
New
|
High
|
Unassigned |
Bug Description
I (and everybody on my team) have a list of 36 bzr repositories we pull in as source deps of our project. As updating them all is very slow, mostly due to network latency, we update them in parallel, 10 at a time, using xargs. Something like
$ xargs -n1 -P10 ./update-repo.sh < sourcedeps.txt
where update-repo.sh does cd $1 && bzr pull, and bzr uses ssh-agent to get the ssh keys, and this is where you come in :) because since we moved to jaunty, we're killing ssh-agent every single time we do this. Not actually killing it; it simply stops responding: suddenly we get these bzrs asking for the passphrase (and failing, obviously). And,
$ ssh-add
Could not open a connection to your authentication agent.
but ssh-agent is still there,
$ ps -C ssh-agent www
PID TTY STAT TIME COMMAND
3208 ? Ss 0:00 /usr/bin/ssh-agent /usr/bin/gpg-agent --daemon --sh --write-
Thanks,
tags: | added: regression-potential |
Changed in seahorse (Ubuntu): | |
importance: | Undecided → High |
21:21 <rmcbride> Anyone seen any behavior like https:/ /bugs.edge. launchpad. net/ubuntu/ +source/ openssh/ +bug/352154 ? A bunch of my team have this as a Nemisis since upgrading to Jaunty Beta BLAH/agent. BLAH gk1OMb/ socket. ssh
21:23 <cjwatson> rmcbride: very likely not openssh's fault at all, but due to seahorse
21:23 <cjwatson> rmcbride: echo $SSH_AUTH_SOCK
21:23 <cjwatson> it's only openssh if it's something like /tmp/ssh-
21:24 <rmcbride> cjwatson: ah, yea. OK that makes sense. One of our Devs spotted it first and wrote the bug based on observed behavior. I'm just following up because I'm seeing it on my dev system now.
21:24 <cjwatson> rmcbride: do actually do that $SSH_AUTH_SOCK check so that we can confirm though
21:25 <rmcbride> /tmp/keyring-
Definitely not openssh's fault, then; I think seahorse is a likely culprit.