Here's my conceptual theory on how to do a decent backup in LPAR mode
without involving the operators.
What I need to do is be able to get all the file systems unmounted,
leaving only root, and then mount that RO, or mount everything RO.
Whether I go to runlevel 1 to do that is somethign of a debate.
If I were to modify runlevel 5, say, to act similarly to runlevel 1, but
keeping the network layer, I could then run a script that does the
unmount/remount process to RO and use FTP to OS/390 to submit a backup job
to JES2. Failing that I would just have the scheduling system run the job
to do the backup when the Linux is scheduled to be down.
I have a DB2CONNECT problem to resolve, where I have to do a manual
db2start after logging on as db2inst1 because a component is not starting -
but that's not important at the moment.
What I do not know how to do, is complex unix scripting. If anyone has some
examples of what I'm trying to do so that I could examine them and see if I
can use that as a bootstrap to get where I need to go?
|---------+-------------------------------->
| | "Ledbetter, Scott E" |
| | <[EMAIL PROTECTED]|
| | TORTEK.COM> |
| | Sent by: Linux on 390|
| | Port |
| | <[EMAIL PROTECTED]|
| | EDU> |
| | |
| | |
| | 09/04/2003 04:02 PM |
| | Please respond to |
| | Linux on 390 Port |
| | |
|---------+-------------------------------->
>------------------------------------------------------------------------------------------------------------------------------|
|
|
| To: [EMAIL PROTECTED]
|
| cc:
|
| Subject: Re: CDL backup and copying the IPL record
|
>------------------------------------------------------------------------------------------------------------------------------|
DFDSS (OK, I know, it is DFSMSdss), should work if you don't mind putting
the CDL volume online to OS390 to dump it to tape. I'd be a little worried
about the state of a Linux file system dumped from OS/390 unless it was
cleanly unmounted or mounted R/O to Linux at the time. Make that a lot
worried. A lot can happen in the time it takes to dump a volume to tape.
You can use the DUMPCONDITIONING option on DSS Copy
(http://www-1.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/Flash10092) to
allow you to copy the volume to another using SnapShot(RVA/SVA) or
FlashCopy
(if your control unit has these features), and dump it from there, while
maintaining the illusion that the volume was being dumped directly from the
original volume. This helps on a restore at the DR site, since you don't
have to reclip the volume after the restore to get the VOLSER to match the
original. Or, use FDR to dump or copy the volume while it is offline, or
use
a SnapShot job to copy the volume while it is offline to OS/390. The
advantage of using a fast replication(Snap, Flash, etc.) utility is that
the
window of opportunity for a file system inconsistency is smaller. However,
inconsistency it is still a big issue, since Linux caches writes in
processor memory. You can use the sync command from the Linux system to
flush the buffers down to the disk, but it is an asynchronous operation so
you don't know when it is complete, and nothing says there won't be another
write cached in processor memory immediately after the sync. Unmounting is
the safest route. Assuming you are talking about your relatively static
Linux OS volume(s), you might be OK.
At any rate, with the correct parameters to invoke a PHYSICAL copy or dump,
DFDSS, FDR, or SnapShot (for volume copy) will all carry the IPL records
for
a CDL volume.
Scott Ledbetter
StorageTek
> -----Original Message-----
> From: James Melin [mailto:[EMAIL PROTECTED]
> Sent: September 04, 2003 2:12 PM
> To: [EMAIL PROTECTED]
> Subject: Re: CDL backup and copying the IPL record
>
>
> Oy.
>
> Well then, how would you recommend this be done? I have no
> decent facility
> at disaster recovery to IPL a recovery CD and mount the disk images to
> that. The vendor is reluctant to load anything from CD at the
> HMC. I'm
> basically trying to take the file system images on the CDL
> volumes and back
> them up using OS/390 tools and then restore them with the
> same tools, to
> diff volumes, map those volumes to the same device addresses
> as they were
> on the original system in VM and IPL.
>
>
>
>
> |---------+---------------------------->
> | | "Post, Mark K" |
> | | <[EMAIL PROTECTED]|
> | | m> |
> | | Sent by: Linux on|
> | | 390 Port |
> | | <[EMAIL PROTECTED]|
> | | IST.EDU> |
> | | |
> | | |
> | | 09/04/2003 02:37 |
> | | PM |
> | | Please respond to|
> | | Linux on 390 Port|
> | | |
> |---------+---------------------------->
>
> >-------------------------------------------------------------
> -----------------------------------------------------------------|
> |
> |
> | To: [EMAIL PROTECTED]
> |
> | cc:
> |
> | Subject: Re: CDL backup and copying the IPL record
> |
>
> >-------------------------------------------------------------
> -----------------------------------------------------------------|
>
>
>
>
> Jim,
>
> I don't think those two statements are either necessarily correct, or
> incorrect. I would say that the "trick" is that the first
> cylinder to be
> dumped to tape (and restored to disk at real cylinder 0)
> needs to be what
> the Linux/390 system thinks is cylinder 0. That's where the
> IPL record
> information is going to be. So, depending on how the
> minidisk is set up
> (if
> using z/VM), and how you dump it, etc., etc., etc.
>
>
> Mark Post
>
> -----Original Message-----
> From: James Melin [mailto:[EMAIL PROTECTED]
> Sent: Thursday, September 04, 2003 1:28 PM
> To: [EMAIL PROTECTED]
> Subject: CDL backup and copying the IPL record
>
>
> If I understand things correctly, a DD dump of a linux volume
> will not copy
> the IPL record, whereas a disk copy of the CDL image on the
> OS/390 side,
> will? Is that a true statement?
> Would this allow you to restore say a slightly pregnant mod-9
> device (our
> dasd person gave me an approximately correct size mod-9 and
> our disaster
> recovery vendor uses real mod-9 geometry 1014 tracks too
> small) to a mod
> 27, and be able to read an IPL record off that device?
>
