Fog client weirdness
-
Two log files here - the first is when I had restarted the FOG service (left side) and the second is after a restart (right side). It just makes no sense to me as to why errors are being returned like this. The only difference between hardware types in our rooms is the presence of an SSD in these labs that has never caused an issue before!
[url=“/_imported_xf_attachments/1/1492_worksdoesnt1.png?:”]worksdoesnt1.png[/url][url=“/_imported_xf_attachments/1/1493_worksdoesnt2.png?:”]worksdoesnt2.png[/url]
-
After trying some other PCs, I have a feeling that this might be to do with the speed at which the PCs using SSDs boot up. If anyone wants to correct me on that for sure, please do - but what I’ll do is reimage one HDD and one SSD machine to see if they are behaving the same way and then secondly turn the FOG service into a delayed start service first and then try and reimage some machines to see if that changes anything.
-
what version of fog are you running?
do you have vmware/virtualbox etc installed on these lab computers? -
FOG 1.2
All our PCs have VMWare Workstation installed (hence the two additional MAC addresses). -
and since it was installed on the image you deployed, the randomized mac address on the virtual network adapter is identical. multiple hosts sharing the same mac address results in the server saying it’s invalid and not delivering tasks. this has been corrected for in the SVN. if you’re willing to update to the latest dev version, this problem will almost certainly go away.
(make a backup and be ready to revert back in case you have problems, the dev version is mostly stable but can have issues, it IS pre-release software) -
I’ll try and update to it - is it going to eventually be a part of the next FOG update then?
And what do you think would make this issue present on only some machines (which happen to all be machines with SSDs)? I never accepted any of the pending MAC addresses (as this did give an issue in 0.32 when I tried to accept one, once) and unless the machine boots from an SSD, it works fine.
Thanks!
-
yes, this will be part of fog 1.3.0
i have never noticed any difference in behavior between computers with HDDs and SSDs so long as they image properly. (i have heard of some SSDs not being recognized and being unable to be imaged, but that has been a very rare issue)
i also noticed that your hardware network card is coming up as “local area connection 2”
is there more then one physical network adapter in these machines? -
On most of our PCs there are, yes - DQ77MK motherboards have an Intel AMT management port and we avoid using this. I’ll reply back here after I try out the delayed start works as I thought it might - but I don’t think having an SSD on the system should make any difference at all. But… it does seem to. Somehow.
-
it looks to me like the vmware mac isn’t the issue then, since you say you’re aware of the related issues and you’ve never approved a pending mac. the issue seems to be that the wrong network adapter is registered with fog.
-
Yeah we know about it and have worked around it with no hassle - actually that has been a way to determine if someone has plugged a machine into the wrong port and we have actually started disabling the second network controller in bios often. I am pretty certain that the adapter isn’t the issue (if it was, it wouldn’t work in either of the cases)
But the key point is that with the same image, on essentially the same hardware, the fog service acts differently - seemingly down to whether or not it uses an SSD. On the SSD machine, the fog service gives the errors - but if you restart the service (stop and start), it works fine.
I just tried to set it to “Automatic (Delayed Start)” and - bizarrely - it is now working fine for those machines that I have done. Could it possibly be that some service is being relied on by FOG that hasn’t started at the point which it is needed? Its really all I can think of - and that could be down to my individual build ultimately. But it isn’t something we have ever seen as an issue before.
-
what version of windows is installed on these machines? if it is windows 8, then it could actually be somewhat related to the SSD. windows 8, i am told, says network adapters are disabled unless it can detect a network attached. that means we can’t get the mac address for them. the current fog client only polls the hardware once on startup, and it requests all active network adapters. if the system is starting up and the fog service starts before the network adapter has initialized, that could possibly be causing the issue.
-
Windows 7. But this sounds kind of logical…
-
windows 7 doesn’t have the same issue. what power saving features do you have enabled on these computers? and does wol work? if the onboard network card is getting set to a low power mode, that could be related
-
Un-ethically, we have everything set to not power down or use power saving modes currently. WOL might work but for the last year or so its been put on the backburner because getting it to work with systems we have limited control over was proving a hassle.
-
have you explicitly disabled the various power states? from the factory, we’ve had some computers set to disable onboard network cards to the extent that they don’t even provide link lights in the off state.
we too have our computers set to not power down or use power saving modes. they just aren’t as reliable that way, in my experience. (although, we have a couple university owned wind turbines, that counteracts the un-ethical-ness, right?) -
Works fine after doing delayed start.
But this kind of tallies with this:
[QUOTE][FONT=sans-serif][COLOR=#000000]One of the first things we try to do in mStart is to sleep for a fair amount of time if the process we are doing does not need to run right at service startup. [B]The idea behind this is that the Windows startup process can be very slow in the first place, so we don’t want the fog service to break the proverbial “Camel’s Back”.[/B][/COLOR][/FONT][/QUOTE]
[url]http://www.fogproject.org/wiki/index.php/Creating_Custom_FOG_Service_Modules[/url]
Im convinced its to do with the speed of services starting up, affected by the SSD in a system. If anyone disagrees, suggest an alternative
-
You are correct, the client currently being distributed is set to start once a core windows service does. However, this is actually too soon on faster systems as not all programs are fully initialized yet. The delayed start is an acceptable solution, or you could link the service to start when a different windows services does. This issue should be resolved with the new client when it is released
-
Ah cool, there hasn’t been a client update in a long time, right? What’s the plan for it?
