attic
attic copied to clipboard
jborg
ssh:// after network re-connection and SIGINT issues
Hello, I did some tests with attic and I am surprised about the features and the performance, thanks a lot! During some tests through ssh:// I tried to disconnect the cable or wireless for a few minutes and I noticed the following issues:
- (generally) after an ssh timeout (i suppose) attic client re-open the connection and run a new attic serve as follow
sshd: pippo [priv] sshd: pippo@notty /usr/bin/python3 /usr/bin/attic serve sshd: pippo [priv] sshd: pippo@notty /usr/bin/python3 /usr/bin/attic serve
and sometimes restart the sync sometimes not. The old one hang and should killed by hand.
- SIGINT implementation doesn't work; I suspect that a thread (ssh?) doesn't receive the signal.
note: I use freebsd 10.1-RELEASE as client and debia 8 as server.
Someone has noticed the same issues? thanks
The locking is not trivial and timeouts of SSH can leave a zombie attic process. Generally on an interrupted connection you want to make sure the attic process is done writing and then it can be killed.
In the fork 'borg' or attic, the locking mechanism was improved. If by any chance you get around testing it, your setup might work more reliable. If you still encounter issues, let us know!
why do you refer to locking? the original post had nothing about it.
His observation was that attic re-spawned another process. To be honest, I did not know that there was such an behavior nor did it ever happen to me. But his ssh log indicates that something had restarted attic.
With a new attic process, the locking should kick in if the old process is still around and busy writing something.
What different theory would you have?