| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 7452.1 | That depends on what you mean by "file"... | WTFN::SCALES | Despair is appropriate and inevitable. | Mon Oct 14 1996 15:37 | 11 |
| 7452.2 | Simple fstat() example | DECC::SULLIVAN | Jeff Sullivan | Mon Oct 14 1996 15:52 | 37 |
| 7452.3 | Thanks | TELEM::VIJAY | | Mon Oct 14 1996 16:48 | 1 |
| 7452.4 | | VAXCPU::michaud | Jeff Michaud - ObjectBroker | Fri Dec 27 1996 19:30 | 3 |
| 7452.5 | Does someone have /a/b/c/d open for write | CXXC::REINIG | This too shall change | Tue Apr 01 1997 20:24 | 5 |
| Is there someway to find out if another process has a file open for write
given just the pathname for the file? I control the other process so if it
requires that the other process do something special, that's fine.
August G. Reinig
|
| 7452.6 | | KITCHE::schott | Eric R. Schott USG Product Management | Tue Apr 01 1997 21:04 | 5 |
| You might start with file locking in both processes.
If that is not good enough, you may have to go to special lock files
to mark more specific information.
|
| 7452.7 | Try lsof ... | NETRIX::"[email protected]" | Detlef Schmier | Wed Apr 02 1997 02:57 | 21 |
| There is also a public domain program called lsof. It shows you every
open file. And you have lots of options.
You can find the latest version in
ftp://vic.cc.purdue.edu/pub/tools/unix/lsof
Detlef.
========================================================================
Detlef Schmier @RTO Digital Equipment GmbH
SBU, OEM/GY Freischuetzstrasse 91
Field Application Engineer D-81927 Muenchen
Pub.Tel. +49-(0)89-9591 2752 DTN 865-2752
Mobile +49-(0)171-3357582
Fax # +49-(0)89-9591 1278 DTN 865-1278
========================================================================
eMail : [email protected]
URL : http://www.digital.de/infocenter/toem-gy/detlef.html
========================================================================
[Posted by WWW Notes gateway]
|
| 7452.8 | | CXXC::REINIG | This too shall change | Wed Apr 02 1997 10:23 | 5 |
| > You might start with file locking in both processes.
Where do I find out more information?
|
| 7452.9 | One possible solution | CXXC::REINIG | This too shall change | Wed Apr 02 1997 12:59 | 14 |
| I think the following will work. The first process will open the file for
write. Every second it will update the access and modification times of
the file using utime. When the second process believes that the first has
crashed (and no longer has the file open for write), it will monitor the
file for a few seconds using stat. If the access and mmodification times
do not change the first process can't have the file open. If the times do
change the first process does have the file open.
The first process and the second process will be on different machines.
The file will be common to the two machines through ASE. How long will it
take for the effects of a call to utime from the first process to be
visible to the second process. I'm willing to wait quite awhile (minutes).
August
|
| 7452.10 | whoa | USCTR1::ASCHER | Dave Ascher | Thu Apr 03 1997 16:10 | 13 |
| re: <<< Note 7452.9 by CXXC::REINIG "This too shall change" >>>
-< One possible solution >-
The first process and the second process will be on different machines.
The file will be common to the two machines through ASE. How long will it
take for the effects of a call to utime from the first process to be
visible to the second process. I'm willing to wait quite awhile (minutes).
Please let me be the first to inform you that files don't get
to be common through ASE. If you have processes on different
machines accessing the same file then they had better be using
NFS for access.
|
| 7452.11 | | DECCXX::REINIG | This too shall change | Sun Apr 06 1997 22:42 | 12 |
| > Please let me be the first to inform you that files don't get
> to be common through ASE. If you have processes on different
> machines accessing the same file then they had better be using
> NFS for access.
While I can use NFS, I thought that the whole point of ASE was to have
two systems connected to the same SCSI bus looking at the same disks.
How can the files not be common? They are on the disk visible to both
machines.
August
|
| 7452.12 | Only VMS does genuine dual-access SCSI clustering? | BBPBV1::WALLACE | john wallace @ bbp. +44 860 675093 | Mon Apr 07 1997 06:28 | 5 |
| But are both hosts able to see the same disks "at the same time" via
SCSI ? I thought only VMSclusters gave you that, and ASE and NT
clusters give you failover between alternate hosts. With ASE and NT you
have to interpose some kind of file service to make failover more
transparent. I could be wrong...
|
| 7452.13 | Failover and load balancing, but not OpenVMS shared disks | UNIFIX::HARRIS | Juggling has its ups and downs | Mon Apr 07 1997 10:03 | 25 |
| ASE does _NOT_ have OpenVMS Cluster style sharing. In a stable
environment only 1 system has control over a subset of the disks. All
other systems in the ASE environment use NFS to access that subset of
disks on the shared SCSI.
The unique feature of ASE is that the an NFS service has its _OWN_ IP
address, so that you can fail the service over to a different node in
the ASE system and the NFS users do not know the difference (since NFS
is stateless).
Multiple NFS services can exist in an ASE system so that you can load
balance the different systems.
Also ASE provides other services besides NFS which also get their own
IP addresses so that they to can be failed over (although if the
connections have any state information, the failover will not be
transparent).
But ASE is _NOT_ OpenVMS clusters.
I do not know what the future holds for TruClusters. Maybe they will
introduce shared disks using the Distributed Lock Manager in a future
release.
Bob Harris
|
| 7452.14 | Thanks for the information | CXXC::REINIG | This too shall change | Mon Apr 07 1997 10:48 | 9 |
| Oh. So even though system two has a direct connection to the disk it
still has to go over the net to system one to get the bits from the disk.
The only time system two can access the bits directly is when system
one has crashed.
Still, I don't see that this affects my proposed algorithm. There is
an NFS connection between the two machines.
August
|