Acronis Backup Advanced 11.7 - some issues with backups of workstations
Hello altogehter,
I have some weird Issues.
First the infrastructure:
- AMS
- Storage Node
- 4 Workstations with Windows Agents
All software is in the most recent version.
On the AMS is a central backup plan that backups all workstations once a week (sunday 1:00 am) with a full backup.
The plan is setup to wake the workstations by WOL.
The Archive is stored in a managed vault on the Storage Node and is replicated to a second managed vault on the same Storage Node.
The 2 managed vaults reside on 2 different NAS.
a) Modify a Backup Plan when workstation is offline
I modified the job on friday.
At this time only 2 of the 4 workstations were running.
During the plan execution this night only 2 of the machines were backed up : the 2 that were online at the moment of plan modification.
The other 2 machines did not run the backup.
Is this a normal behaviour?
From my point of view it is quite normal that not all machines are running if the plan is modified and I assumed that the plan configuration is transferred to the agents (that were offline in the moment of modification) if it connects the next time.
b) WOL only at scheduled time not with manual start
This morning I started all the workstations and did some modifications in the backup plan (I changes the comment) to force an update to the workstations.
Then I shutdown the workstations to test the plan.
In the plan overview I tried to start the plan.....no success.
I then started the workstations by sending a WOL packet...they woke up and...the plan started.
Is this a bug or a feature ? ;-)
c) plan is not finished if valildation is configured
In one of the previous setups I configured a validation after the full backup.
I also configured a shutdown of the workstation if the backup is finished.
My assumption was, that the storage node executes the validation task and there is no problem to shutdown the workstation.
But his is not correct!
I looked into the logs and saw some entries that shoed:
- the agent delegates the validation task to the storage node --> I assumed this
- the workstation is shutdown --> yes I configured it to do so
- the storage node exectues the validation and finishes --> perfect
- there is a task on the agent machine still running --> HUH ??
I asssume that the agent on the workstation delegates the validation task, but needs a response after the validation in the mean of "yes I finised the task you delegated to me".
Is this assumption correct and how can I fix it?
Are scheduled plans deleayed until this handshake between storage node and agent is done?
Currently my backup plan has no validation task.....
e) schedule validation task
I cannot find a possibility to schedule a validation outside a backup plan.
Is that correct?
The only thing I see is that I can create a time plan for validation within a task.
But who executes this validation ?
Only the storage node or is there a need of communication with the agent?
f) replication to second archive also needs agent
In the logs of the backup plan I also see some entries during the replication/retention tasks to the 2ns location that seem to need communication with the storage node.
The result is, that the complete backup plan is not finished until the worstation is online again (I assume that there is also some kind of handshake).
Is there a workaround to solve this?
Greets
Sven
Hello Fedor,
thanks for the reply (I hope Fedor is your first name...).
a) I will keep an eye on it and come byck if I have an idea to reproduce it.
b) Okay....it´s noted in the documentation - I overlooked it.
But there are so many information I cannot read all ;-)
If I know this limitation it´s okay for me as I can address it....
c) The planned improvement is perfect.
It´s a limitation that limits a lot in case of workstation backups.
I think it´s not only me who initiates a shutdown on the workstations to save power.
And if the agent still needs some response from the storage node it´s quite ugly.
d) ups I missed it in my first post ,-)
e) Okay I will have a look this evening.
This would be a workaround for c)
f) Replication is a cool feature, but I think the implementation can afford some adjustements ;-)
Replication should either be done by the ASN on it´s own (I think all relevant data are accessible by the storage node without the need of a running agent) or the monitoring scheme should be altered a bit.
One question:
If a scheduled backup is in the state described of f) (so to speak waiting for the agent´s machine to come online againto finish the monitoring handshake) does this delay a later scheduled backups until all workstations are online again and finish the monitoring?
This would be quite a big issue.
S.
Regarding the last question - no, agents work independently and launch their tasks regardless of what happens on other agents
Hello Fedor,
just to understand you correct:
The following scenario:
- all plans are central plans on a Win2012 Server
- all plans store the backup in a centralized vault on a NAS and the archive is then replicated to a second centalized vault on another NAS
I have the following backup plans:
P1
- scheduled at 1:00 am each sunday
- full backup of Workstation 1
- Workstation is started by WOL and then shutdown after Backup (thats the reason why the backup plan is not finished as the handshake between storage node and agent on the workstation is not finished as described in the first post=).
P2
- basically same as P1 but for Workstation 2 and scheduled on 1:05 am each sunday
P3
- basically same as P1 but for Workstation 3 and scheduled on 1:10 am each sunday
P4
- basically same as P1 but for Workstation 4 and scheduled on 1:15 am each sunday
P5
- scheduled each day 2:00 am
- GFS-backup of a fileserver
- no WOL needed and no shutdown
The slightly different start times should bring the backups in a sequence and all backups should run back-to-back.
That is the idea....but does it work ???
Does P2 start although P1 is not finished ?
Regards
Sven
Hey,
yesterday I did a quick check and was really astonished.
I scheduled the backup jobs in the way described in #4.
The intention of this test was
- test if sequencing works
- test if the different backup plans are delayed by the previous one.
First of all the good news: Yes, all backup plans are executed and a hanging plan (because of the not finished handshake) has no effect. perfect....
Now a few points I´m surprised:
a) Sequence
The 4 backup jobs had start times in a distance of 5 minutes.
My assumption was, that a backup plan starts if the previous one is finished...but NO! ;-)
They start at their planned times with their job and the 4 backup plans run in parallel.
But why ?!
In the documentation is stated that a task is waiting if the needes resources are used by another task.
So let´s check:
- the backup source is separate for each backup plan --> okay
- the stroage node is a common object --> okay but it is built for it....so okay also no locking resource
- all backups in the same vault --> here I´m wondering
I thought that different backups that use the same vault cannot run on the same time.
Or is the resource in the meaning of the delay-a-backup-plan issue an archive or backup?
I monitored the status of the different tasks and the catalog creation was not run in parallel but in sequence.....
So can ACRONIS provide a list of resources that lead to a lock and so a delay in a later backup plan or task if they are already used?
Hmm back to the question of sequencing:
As stated, all jobs use the same vault and so the same HDDs.
My intention of scheduling was to prevent concurrent access on different archives to prevent HDD head banging.
How does the storage node handles this situation that it must write to different files on the same target?
Is there any mean to prevent performance drop?
b) no running activity
After all plans are finished the backup-plan view showed all plans OK but no actual finishing time.
If I then clicked the button in the lower right corner og backup-plan view "current activities" (in german its "aktuelle Aktivitäten") I see that the jobs are still runnign that need the handshake with the workstation´s client.
Is this by incident or accident ?
Regards
Sven
Plans wait for other plans only on the same machine. Up to 10 machines can back up at the same time to a storage node ( https://kb.acronis.com/content/2929 )