[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: does killer kill ssh sessions? / Re: Strange LTSP phenomenon: Clients sessions freeze



Am Dienstag 22 September 2009 schrieb Petter Reinholdtsen:
> This is how ps behave for all users with usernames longer than 8
> characters -> only the uid is shown.  Other tools like top behave the
> same way, and this is part of the reason I strongly recommend using
> max 8 characters in usernames.

Okay, thank you for this clarification.
>
> When that is said, I agree that killer should be changed to kope with
> it.  Anyone able to provide a patch?

Not me, so far. I read the script carefully, and understood much of it - 
but (still verbose) commentary didn't mention "anonymous". Is my 
impression right that first the table of processes in question is filled 
and then entries are removed if matchin certain aspects?

Then comparing uid and ruid might be the section were anonymous 
processes are killed? (line?)

Proposal (@BTS): introduce option -a/--ignore-anonymous

From yesterday, I found several open sessions left behind (7 characters 
in login name), including soffice zombies.

killer triggers to "sh" here as well, while soffice child processes are 
left behind. Maybe the output of "killer -d" gives hints here:

        vhs2-32  vhs2-32  11046 11046 ?        17442 17440   0 sshd           
        vhs2-32  vhs2-32  11046 11046 ?        17447 17442   0 bash           
        vhs2-32  vhs2-32  11046 11046 ?        17450 17447   0 ck-
launch-sessi
        vhs2-32  vhs2-32  11046 11046 ?        17930 17450   0 ssh-agent      
        vhs2-32  vhs2-32  11046 11046 ?        17938 17450   0 x-
session-manag
        root     vhs2-32      0 11046 ?        17972     1   0 
start_kdeinit  
        vhs2-32  vhs2-32  11046 11046 ?        17973     1   0 kdeinit        
        vhs2-32  vhs2-32  11046 11046 ?        17976     1   0 
dcopserver     
        vhs2-32  vhs2-32  11046 11046 ?        17979 17973   0 klauncher      
        vhs2-32  vhs2-32  11046 11046 ?        17981     1   0 kded           
        vhs2-32  vhs2-32  11046 11046 ?        17983     1   0 
gam_server     
        vhs2-32  vhs2-32  11046 11046 ?        18007 17938   0 kwrapper       
        vhs2-32  vhs2-32  11046 11046 ?        18009     1   0 ksmserver      
        vhs2-32  vhs2-32  11046 11046 ?        18010 17973   0 kwin           
        vhs2-32  vhs2-32  11046 11046 ?        18012     1   0 kdesktop       
        vhs2-32  vhs2-32  11046 11046 ?        18018     1   0 kicker         
        vhs2-32  vhs2-32  11046 11046 ?        18019 17973   0 kio_file       
        vhs2-32  vhs2-32  11046 11046 ?        18021     1   0 kaccess        
        vhs2-32  vhs2-32  11046 11046 ?        18035 17973   0 artsd          
        vhs2-32  vhs2-32  11046 11046 ?        18041     1   0 knotify        
        vhs2-32  vhs2-32  11046 11046 ?        18043     1   0 
kpowersave     
        vhs2-32  vhs2-32  11046 11046 ?        18045     1   0 klipper        
        vhs2-32  vhs2-32  11046 11046 ?        18047     1   0 korgac         
        vhs2-32  vhs2-32  11046 11046 ?        19016 17973   0 soffice        
        vhs2-32  vhs2-32  11046 11046 ?        19208 19016   4 
soffice.bin    


After "killer -d" this process is left behind:
> vhs2-32  19208  0.2  3.9 517304 159024 ?       SNl  Sep26   4:05 
/opt/openoffice.org3/program/soffice.bin -writer


Reply to: