| 1) In the server, how to detect that no recover, save, savefs
savegrp or media command is running. According to the man
pages, a "ansrd" should be started for each, but I only saw
it for recover? I should write a script which has to unmount
the tape, invite the operator to replace the stack by the
new one and then inventory it. How to avoid the collisions?
>> ps axw | egrep 'TIM|nsr|save|recover'
>>
>> naturally, this is a basic thing, but any shell could parse this for
>> particulars, and avoid collision.
2) Is it a way to place the SERVER in a state where it could
still inventory a JUKEBOX, but where it refuses the save
or recover sessions? Maybe, nsrjb itself is locking nsrd ?
>> hmmm, as long as there are no labelled tapes, saves will wait.
>> if there are no active devices, saves wait for a tape. Use
>> nsradmin and edit resources with a script to activate/deactivate
>> devices (man nsradmin) -- oops, deactivating devices, means
>> your jukebox cannot inventory!! sigh.
3) Is the inventory the only way to give the server the
contents of the JUKEBOX? Can I edit the "loaded volumes"
resource of the nsrjb.res file? I tried today, but it works
only if I restart DECnsr. Is it another way to achieve the
same? (a signal sent to a nsr process maybe ?)
>> yes, to my knowledge. *please* do not edit nsrjb.res - you are
>> pointing a loaded gun at your foot with your finger on the
>> trigger...
>> hmm, what do you mean: "give the server the contents" -- are you trying
>> to ascertain what the JB has, or, are you trying to get NetWorker
>> to know what the JB has in it?
|
| But, maybe you mis-read my first question:
1) In the server, how to detect that no recover, save, savefs
^^^^^^^^^^^^ savegrp or media command is running.
I started on a *client* a manual save needing a pool not present in the
jukebox. The server displayed the usual "waiting for a writeable tape
of the pool...". At that time, a "ps -ef | grep 'nsr|recov|sav'" does
NOT show anything.
My problem is the following: If, at that time, I start a script doing
the Inventory of the 7 slots of the jukebox, as soon as a tape of the
needed pool is inventoried, the waiting save will most likely start and
then what about to the inventory process? In other words, I don't know
if the inventory process will complete before the save (or any other
recover/savefs/savegroup operation) starts ?
> 2) Is it a way to place the SERVER in a state where it could
> still inventory a JUKEBOX, but where it refuses the save
> or recover sessions? Maybe, nsrjb itself is locking nsrd ?
Still no thought? Can I change the paralellism of the server to 0 ?
For me this is a important feature that should exist. When a customer
is buying a TLZ877 jukebox, he is NOT buying another tape device to
read/write the tar/cpio/vdump tapes needed for software upgrade..etc.
So, if we were able to freeze momentarily NetWorker but still can use
nsrjb to load tapes (nsrjb -l -n), it would be a nice security feature.
>> hmm, what do you mean: "give the server the contents" -- are you
>> trying to ascertain what the JB has, or, are you trying to get
>> NetWorker to know what the JB has in it?
- To get NetWorker to know what the JB has in it. The inventory process
is long on a TLZ877 and we perfectly know in advance the contents of
our 7 Tape stack canisters. So, I can edit a file or have NetWorker
read a file containing the well-known contents of our 7 canisters, the
long inventory process becomes unnecessary.
|
| There is another note string about using a jukebox to do
tar/cpio/etc. with some safety, but always ALWAYS remember
that NetWorker was NOT designed to share jukeboxes with
outside applications. You do so at your own peril,
and at the peril of your customer's data!
I do not know for certain of anyway to make sure an inventory completes
before a pending save would take off. It irritates me too, but that
is the current state of things. We've requested this change from Legato.
We'll keep you posted.
Kevin
|
| > I do not know for certain of anyway to make sure an inventory completes
> before a pending save would take off. It irritates me too, but that
> is the current state of things. We've requested this change from
> Legato.
This evening, I made a test and it seams to indicate that NetWorker
V4.2B -Patch 12 is waiting the completion of the inventory.
On the server, I started : nsrjb -I -S1-7 (In fact, all slots)
A few second after, on a client I started a save to a POOL whose a tape
was present in the first slot to inventory. The inventory completed (15
minutes) before the first inventoried tapebe mounted and the save
started.
For the rest, I also verified that nsrjb is NOT starting any action on
tape if NetWorker is using the drive. (Exit status = 3) and even is
waiting for the completion of another nsrjb started by NetWorker or
another user.
SO, my script seams working as expected.
Thanks for your support.
Manu.
|