3.9SP1 and repair snapshots

[MartinGoldstone-Keele]

Hi,

We are currently in the process of rolling out a new lab of 70 machines, and this morning I've had reports that some of the systems are taking an extremely long time to boot (we are using the Fast Return mode on these). I'm currently sitting next to 3 machines that have so far taken over an hour, and are approx 20% through "Downloading shared files" on the "Checking partition table" segment. The machines are in Clean Start, and according to the Rembo console, are caching their recovery snapshots. Separately, I started another couple of machines, 1 of which took about 20 minutes to boot, the other about 8. After going through the management console, I've discovered that the machines that have taken the longest to boot have 3 repair snapshots stored, the 20 minute machine had 2, and the 8 minute machine only one.

As we only require the machines to have the latest snapshot stored, is there any way to set Net-runna to only store 1 snapshot, or an easy way to delete the surplus snapshots (currently the only way I've seen in the interface to do it is within each clients' boot settings screen). Ordinarily, a 20 minute boot time would not be too much of an issue for us (we would simply use wake on lan to start the machines earlier), but with a boot time which seems like it may be several hours, it seems hardly worth it to shutdown the machines at all.

Additionally, in this lab we've deployed the machines with a filter in the deployment profile to try to prevent anti-virus updates from being reset. Seemingly this has had the effect of not deploying to that directory in the first place. I've only just discovered this issue, so I'm continuing to investigate, but I am wondering if this is expected behaviour or if this has been encountered before.

Hope someone can help with these issues.

Cheers,

Martin

[martin]

Martin,

Just to confirm, you have deployed the clients using a Deployment Profile with the Repair Type set to "Fast Return" and gone through the "First Boot" and "Create Repair Snapshot" phases?

What do you mean by the clients being in "Clean Start" mode? This is a debug level meaning that no plugins are executed during preboot - so shouldn't appear when you are doing a normal boot with repair.

Can you set the debug level for one of the problem PCs to "Verbose", PXE-boot it and send a copy of <mac address>.log plus the net-runna Support Pack for that client to support@net-runna.com.

If you wish to limit the number of Repair Snapshots, in the Management Console take the Manage/Configure Boot Server Synchronization menu option - there are more details in the online Help under "Service Pack 1 Addendum" in the "Boot Servers" (plus the "Multiple Repair Snapshots") section.

We can look into the "missing directory" issue when we fully understand the reasons for the long repair time.

Regards,

Martin

[MartinGoldstone-Keele]

Sorry for the delay in replying.

Yes, they have been deployed as described. I've now switched clean start off (it seems this was set by mistake on the location these machines are in, though I do seem to remember we once had a few machines that refused to work unless they were in clean start mode, but will need to check this to confirm; either way, it is not the case with these machines, so they have been set to no debug information).

I've set verbose mode on a machine that this morning took a long time to boot. I've had some difficulty creating the support pack (several times the management console failed to create it, but eventually succeeded), but the boot log and support pack have been sent to the email address requested. At any rate, it seems at the moment that this long repair time is only happening the first time the machine is repaired after being deployed.

Concerning the multiple repair snapshot issue, I've set the option suggested, and gone through the online help. The multiple repair snapshots section suggests looking at the synchronisation tool under chapter 5, though I cannot find it there. I found some help regarding the synchronisation tool, but although I checked the box to synchronise repair snapshots and ran it by scheduling it, it did not appear to delete any, and no further space was reclaimed by running the sweep shared repository operation. Does this tool only work when synchronising against another server? The only way I have found to delete these so far from within Net-runna has been to do it manually, which is exceedingly time consuming. Additionally, further deployments have once again created additional repair snapshots. Is there a way to prevent this from happening?

Regarding the missing directory issue, I have now returned to using repair all instead of fast return. I did some more testing before switching back, and discovered that the directory is there immediately after deployment, but is then missing after the first repair. I can only guess, but it seems a list of files to repair is built, and this directory is omitted as it is on the exclusions list, and this same list is used when determining which files should be on the drive.

Thanks for your help so far,

Martin

[martin]

Martin,

just to let you know that we are going through the log-files and will try to duplicate some of your environment to see if we can find out what is happening - it seems to be re-building the cache partition at each boot?

One thing that you could try in the meantime is to enable Unicast network transfers for one of the clients (via Boot Settings/PXE Settings) and see if that makes a difference.

Back soon!

Regards,

Martin

[martin]

Martin,

just a couple of other things to check...

What is the load on the Boot Server when you are attempting a deployment or repair? Even relatively simple things seem to take quite a while to complete.
Are you running a server-based antivirus program performing on-demand checking? If you are scanning the repository it will have a massive impact (and not give you much benefit as the BLK files are encoded).

Martin

[martin]

Martin,

going over your log-files from the 25th to 27th January you have deployed images in the form:

<image>-v01 into partitions DIAGS:??% NTFS:??% EMPTY:0% EMPTY:0%
<image>-v02 into partitions EMPTY:0% NTFS:75% EMPTY:0% EMPTY:0%
<image>-v02 into partitions DIAGS:1% NTFS:74% EMPTY:0% EMPTY:0%
<image>-v04 into partitions DIAGS:1% NTFS:74% EMPTY:0% EMPTY:0%

...but with net-runna it is recommended that you avoid diagnostic partitions, as they can have unexpected side effects (because we still try to "understand" and possibly repair them).

You should remove the DIAGS partition using the instructions found in How to remove diagnostics partitions using net-runna and re-deploy one of the clients (preferably using a temporary Deployment Profile so that other machines are not affected) to see if that improves things.

Once we have a clean image, we can investigate further your Repair Snapshot and missing directory issues.

Regards,

Martin