It does not explain why when backing up other type of data I can send index to remote AFTD device, and while backing up console i must have local AFTD device.
Maybe i was not clear in my question. I'll ask ths way: what options must be set to back up index to separate pool (which uses remote storage device) node during Console Server backup?
Console is running on backup server so it will look at server properties. If you wish to get a bit more clever setup, you can create pool called FOO and have in selection criteria:
index:
bootstrap
(pay attention that index has ":" and bootstrap and doesn't). Label device for that to be on backup server as both index and bootstrap should be saved on server. You can take this step further and configure networker not to save indices when group finishes (eg. by setting no index save) and have special index group which saves only index and bootstrap (eg. by setting index only in group properties). This entirely depends on how you wish to have it.
Normally, I would never keep NMC on the same box as server and I think there is even EMC recommendation which says that if you have more than 50 clients you should not do that.
What you are doing is silly - you are backing up ssid from backup server not to backup server device, but somewhere else.
I would suggest that you have setup where:
a) you have minimized number of pools and they are based on retention only (so no need for extra index pool)
b) use no index save
c) data from backup server gets saved on backup server, data directed to storage node gets saved to storage node
d) use index groups to trigger index backups when necessary (depends on your overall schedule or needs like before working hours, during lunch, after working hours or anything else that may work for customer).
e) use separate host for NMC
Now, above is easy with VM and DD (or any other de-dupe disk device solution). With tape is a bit heavier, but again possible, but nowadays disk rules anyway. I would not rely on snapshots of VM for backup server protection - I would rather use index and bootstrap backup as this is intended and supported. If you thought of making setup where only devices are on storage node and backup server is just being snapshotted - I don't think you will get much sympathy for that under current workflow design.
Maybe it is "silly" looking from NW philosophy perspective, but I'd like to minimize effort needed to recover NW server itself.
And treating it as virtual appliance seems to me to be the optimal approach (especially when you leave backup system in hands of vSphere/Hyper-V admin, who does not know much about backups).
I know downside of this situation - stopping, services before snapshot is not very good solution.
What is more - after reinstall NW can scan and recover data from remote device like DD.
So why I must place primary bootstrap backup on disks of "less" reliable virtual machine?
I know it can be cloned, but that is not the case.
And also it is not clearly said in documentation about pool requirements for backup of NW server (and I find it inconsistent after this tests).
NW doesn't come as virtual appliance so you must handle it as regular application. Not sure about your env, but in mine due to high IO I do not plan to protect it via snapshots through hyervisor. Snapshot integration with application exists on Windows platform via writers and there is no such thing for NW - so, for me that is no go. And if you consult support, you will most likely get same answer. NW has rather good mechanism of protecting itself so no one stops you from protecting the box with snapshot, but application itself will still require nsrdr most likely. This may change in 8.5 or later as many things are about to change, but that's not present, but rather future.
I do not agree that VM is less reliable. Actually, I built VM server which is now more powerful and more reliable than my HPUX and AIX power horses were. You choose what fits application when designing solution and not other way around.
ble1
4 Operator
•
14.4K Posts
0
February 17th, 2015 09:00
index should be backed up on server as indexdb belongs to server client.
Tomsmi
17 Posts
0
February 17th, 2015 12:00
It does not explain why when backing up other type of data I can send index to remote AFTD device, and while backing up console i must have local AFTD device.
ble1
4 Operator
•
14.4K Posts
0
February 18th, 2015 23:00
You can configure that for index, but that is not nor ever was recommended.
Tomsmi
17 Posts
0
February 19th, 2015 00:00
Maybe i was not clear in my question. I'll ask ths way: what options must be set to back up index to separate pool (which uses remote storage device) node during Console Server backup?
ble1
4 Operator
•
14.4K Posts
1
February 19th, 2015 05:00
Console is running on backup server so it will look at server properties. If you wish to get a bit more clever setup, you can create pool called FOO and have in selection criteria:
index:
bootstrap
(pay attention that index has ":" and bootstrap and doesn't). Label device for that to be on backup server as both index and bootstrap should be saved on server. You can take this step further and configure networker not to save indices when group finishes (eg. by setting no index save) and have special index group which saves only index and bootstrap (eg. by setting index only in group properties). This entirely depends on how you wish to have it.
Normally, I would never keep NMC on the same box as server and I think there is even EMC recommendation which says that if you have more than 50 clients you should not do that.
ble1
4 Operator
•
14.4K Posts
0
February 20th, 2015 01:00
What you are doing is silly - you are backing up ssid from backup server not to backup server device, but somewhere else.
I would suggest that you have setup where:
a) you have minimized number of pools and they are based on retention only (so no need for extra index pool)
b) use no index save
c) data from backup server gets saved on backup server, data directed to storage node gets saved to storage node
d) use index groups to trigger index backups when necessary (depends on your overall schedule or needs like before working hours, during lunch, after working hours or anything else that may work for customer).
e) use separate host for NMC
Now, above is easy with VM and DD (or any other de-dupe disk device solution). With tape is a bit heavier, but again possible, but nowadays disk rules anyway. I would not rely on snapshots of VM for backup server protection - I would rather use index and bootstrap backup as this is intended and supported. If you thought of making setup where only devices are on storage node and backup server is just being snapshotted - I don't think you will get much sympathy for that under current workflow design.
Tomsmi
17 Posts
0
February 20th, 2015 01:00
I've already tried "No index save" option and it seems to be only way to backup Console on remote node.
Basically I'd like to achieve configuration consistent for all clients and all data types
For regular FS client this setting works:
1. Separate pool for data on remote Storage Node.
2. Separate pool for bootstrap data on local device on netwoker server (save set: bootstrap).
3. Separate Pool for index data on remote Storage Node (save set: index:)
But I do not know why for Console backup separation is allowed for data, but index and bootstrap must be on local NW device.
Those test I was doing ony in lab environment, so there is no real need for this in real production.
But - what if I'd like to treat NetWorker server only as management plane installed in tiny VM (which i can snapshot anytime)?
In that configuration I really do not need any backups stored on that VM.
Thank you for explanation, and confirmation about those restrictions in NW.
Tomsmi
17 Posts
0
February 20th, 2015 02:00
Maybe it is "silly" looking from NW philosophy perspective, but I'd like to minimize effort needed to recover NW server itself.
And treating it as virtual appliance seems to me to be the optimal approach (especially when you leave backup system in hands of vSphere/Hyper-V admin, who does not know much about backups).
I know downside of this situation - stopping, services before snapshot is not very good solution.
What is more - after reinstall NW can scan and recover data from remote device like DD.
So why I must place primary bootstrap backup on disks of "less" reliable virtual machine?
I know it can be cloned, but that is not the case.
And also it is not clearly said in documentation about pool requirements for backup of NW server (and I find it inconsistent after this tests).
ble1
4 Operator
•
14.4K Posts
0
February 20th, 2015 03:00
NW doesn't come as virtual appliance so you must handle it as regular application. Not sure about your env, but in mine due to high IO I do not plan to protect it via snapshots through hyervisor. Snapshot integration with application exists on Windows platform via writers and there is no such thing for NW - so, for me that is no go. And if you consult support, you will most likely get same answer. NW has rather good mechanism of protecting itself so no one stops you from protecting the box with snapshot, but application itself will still require nsrdr most likely. This may change in 8.5 or later as many things are about to change, but that's not present, but rather future.
I do not agree that VM is less reliable. Actually, I built VM server which is now more powerful and more reliable than my HPUX and AIX power horses were. You choose what fits application when designing solution and not other way around.