[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference smurf::ase

Title:ase
Moderator:SMURF::GROSSO
Created:Thu Jul 29 1993
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2114
Total number of notes:7347

2034.0. "Can't relocate a disk service running Usenet News" by NNTPD::"hoveys@zko.dec.com" (Shari Hovey) Wed Apr 30 1997 11:37

Hi,

I am having "intermittent" problems failing a disk service on a two
node ase cluster. There are two disks that are failed over. /ase1
and /ase2. When we first set up the ase cluster, we had no problems
failing it back and forth on the two nodes. It was left alone for 
about four hours. Then we were no longer able to fail the service.
It kept saying that the /ase1 device was busy. I ran lsof on both the
disks and it came back with nothing. The service stop script shutdowns
inn but is unable to unmount the disk, so it restarts inn. I put
a "sleep 5" after inn shutdown to give syslogd a chance to complete
it's writing to the inn log files. Still no luck.

So we rebooted both systems and now we can fail the service back
and forth like there's no tomorrow. I will be using these systems
to train a customer (Hong Kong Telecom) on INN and how great ASE
is for a Internet Service Provider who has LOTS of newsreaders.
I'm afraid that by the time I come in tomorrow morning, the
systems will not want to fail over again.

What else is there to check in situations like this?

Thanks in advance,

Shari


[Posted by WWW Notes gateway]
T.RTitleUserPersonal
Name
DateLines
2034.1COMICS::CORNEJWhat's an Architect?Wed Apr 30 1997 12:526
    Does lsof report anything if a user is "cd"'d to the filesystem?
    
    (I don't know - I've never looked)
    
    Jc
    
2034.2Yes, lsof reports cd'sNNTPD::"hoveys@zko.dec.com"Shari HoveyThu May 01 1997 01:319
Hi,

Yes, lsof will report a process id if someone is cd'd to the disk.
Any other suggestions?

Thanks,

Shari
[Posted by WWW Notes gateway]
2034.3???BRSDVP::DEVOSManu Devos NSIS Brussels 856-7539Thu May 01 1997 10:1210
    If DECsafe is telling the device is busy, then lsof should show you
    which process.
    
    Did you try to place lsof IN YOUR STOP script (at the end) to display
    to the console its output?
    
    Maybe you started asemgr from a shell whose current directory is in the
    /ase1 directory.
    
    Manu.
2034.4lsof does not detect all reasonsSMURF::MYRDALThu May 01 1997 15:1121
    Hi,
    
    I asked this question of one of the filesystem engineers here.  This
    was part of his reponse:
    
    "I'm not convinced lsof is capable of detecting all references.  I know
    that lsof cannot detect mmap references, which will prevent a
    filesystem from being unmounted.
    
    There is/will be a patch to detect mmap references via the fuser
    syscall. Our fuser utility will detect all references.  You may need to 
    build a steel version fuser (statically-linked) to provide to the customer.
    I would recommend they investigate for ref-holders either by using fuser
    (preferably with the patch to detect mmap references), or via the good
    old-fashioned approach of killing processes until they find the guilty
    party."
    
    Just passing on the info ......
    
    -- Greg