Cant deploy images from master node and storage node at the same time.
-
Hello folks!
Running fog version: 1.5.6
OS on master node. Ubuntu 18.04.2 LTSI discovered a new problem with my FOG-Setup. I recently added a storage node to my fog installation. Replication works fine but when i want to deploy an image to a pool of 50 machines the image deploying only works from my storage node. But when I turn off my storage node all of sudden the deployment works also from my master node. So I can only deploy 10 machines at a time and not 20.
I have only one storage group (default) with one master and one node.
I also noticed ( when master and node are running) that in the active task view it shows in the line ‘working with node’ DefaultMember (master node) but when they start the deployment process it changes to my storage node.master node:
storage node:
I don’t know if it has something to do with this error described here https://forums.fogproject.org/topic/10006/ubuntu-is-fog-s-enemy
Because I fixed it recently with the tutorial Tom wrote in this thread.Well thank you for your help!
I appreciate it very much. -
So just to be clear on the 11th simultaneous deployment the master node is not used? I don’t think the scheduler is very smart. It doesn’t load balance the deployments but uses the overflow method of workload distribution.
Slightly off topic, unless your fog server is running on a 10GbE network, having more than 3 simultaneous deployments will saturate a single 1GbE network link. If you are trying to deploy to 20 systems at the same time, you might want to consider multicasting if your infrastructure is capable of multicasting. Understand this is a networking decision not a FOG decision.
-
@george1421 said in Cant deploy images from master node and storage node at the same time.:
So just to be clear on the 11th simultaneous deployment the master node is not used? I don’t think the scheduler is very smart. It doesn’t load balance the deployments but uses the overflow method of workload distribution.
Yes, when the 11th client checks in it queues up and waits that a slot on the storage node becomes available. What do you mean with the scheduler is not working smart? Is it something i can configure?
Now the dashboard shows on the ‘Storage Node Disk Usage’ view in the drop down box again my storage node. I think it wasnt visible the last time I checked. But I didnt do anything as far as I know.
Unfortunately I can’t test it right now because I discovered that in Uefi mode Windows changes my boot order and I have to manage that problem first to test it again.Slightly off topic, unless your fog server is running on a 10GbE network, having more than 3 simultaneous deployments will saturate a single 1GbE network link. If you are trying to deploy to 20 systems at the same time, you might want to consider multicasting if your infrastructure is capable of multicasting. Understand this is a networking decision not a FOG decision.
Yeah I didnt think about it. Its maybe something worth considering. But the network part isn’t in my hand right now so will ask my coworker about it.
-
@paat said in Cant deploy images from master node and storage node at the same time.:
I discovered that in Uefi mode Windows changes my boot order
Yes MS Windows will do that for you to help you cause you don’t know what you really want.
There are two schools of thought here.
- Don’t have/use boot through PXE but use the boot menu when you specifically want to pxe boot the target computer (that is what I do on my campus because I want the techs to “know” what system they are imaging). That why when windows does a feature update it doesn’t fix it again and again for you.
- Use a bios/firmware tool like Dell’s CCTK to change the boot order in the setupcomplete.cmd batch file once windows WinSetup/OOBE has finished.
-
@george1421 said in Cant deploy images from master node and storage node at the same time.:
@paat said in Cant deploy images from master node and storage node at the same time.:
I discovered that in Uefi mode Windows changes my boot order
Yes MS Windows will do that for you to help you cause you don’t know what you really want.
There are two schools of thought here.
- Don’t have/use boot through PXE but use the boot menu when you specifically want to pxe boot the target computer (that is what I do on my campus because I want the techs to “know” what system they are imaging). That why when windows does a feature update it doesn’t fix it again and again for you.
- Use a bios/firmware tool like Dell’s CCTK to change the boot order in the setupcomplete.cmd batch file once windows WinSetup/OOBE has finished.
- But do i get this right if you go this way you have no automatism?
- Yeah I’ve looked into this and I think this is the way to go for me. I’ll setup a basic Windows install (for test purposes) tomorrow at work to get along with this boot mess and can look again into the actual master/storage node problem. Then I’ll report back!
Thank you for you thoughts.
-
@paat said in Cant deploy images from master node and storage node at the same time.:
- But do i get this right if you go this way you have no automatism?
Yes you won’t have unattended deployments if you don’t pxe boot through fog, BUT you won’t accidentally wipe the CEO’s laptop either if you force the IT tech to at least sit in front of the computer to start imaging process. The rest of this method is all no toch.
-
@george1421 said in Cant deploy images from master node and storage node at the same time.:
@paat said in Cant deploy images from master node and storage node at the same time.:
- But do i get this right if you go this way you have no automatism?
Yes you won’t have unattended deployments if you don’t pxe boot through fog, BUT you won’t accidentally wipe the CEO’s laptop either if you force the IT tech to at least sit in front of the computer to start imaging process. The rest of this method is all no toch.
Hey I created a new image with a script which starts with the setupcomplete.cmd and this fixed the Uefi-Boot issues via bcdedit.
I didn’t use the dell cctk tool yet because I have still a few Fujitsu machines left but maybe i’ll switch to it because it seems very powerful (firmware deployment, bios managment etc.)The actual problem seems to solve by itself. I don’t know what happened exactly but I deployed the image to 50 machines and both servers (master and node) deployed the images. Maybe I was to impatient.
If I’m right the master server checks if there are open slots on each server and distributes and controls the image process
I think the thread can marked as solved for the moment.Thank you for your help George!
