Discrepancies between an HP blade memory and what Proxmox sees

[philjans]

Hi,
I have an Dell blade server rack and I need more memory on one of my bland but I found that it has 24gb ram but somehow Proxmox sees 55gb...


Where could that be from.
I want to buy and ugprade the ram level but I was not expecting 24gb...

Also, I read something that I am affraid would apply here: "I am using XenServer hypervisor and all these 4 servers are in a pool. I heard that, all servers in the pool should be in identical configuration by cpu, memory and all hardware configuraiton."
Is this applies to Proxmox also?

Thanks for the info

 

[apoc]

Your KSM-Sharing Vvlue is very high.
https://pve.proxmox.com/wiki/Dynamic_Memory_Management
Maybe that is related to your observation?
I can't tell how PVE actually calculates these values ...

 

[dcsapak]

pve just takes the info from the kernel.. what does 'free -h' show? can you post the output from 'dmesg' ?

 

[philjans]

pve just takes the info from the kernel.. what does 'free -h' show? can you post the output from 'dmesg' ?

Hi,
Here's what free -h shows:

And I have attach the dmesg logs
Thanks you!

 

[dcsapak]

ok the dmesg output is incomplete, can you post the output of

journalctl -b -t kernel

?

 

[philjans]

Hi,
How about this one:
tx!

 

[dcsapak]

it seems there went something wrong with the output. can you redirect the command into a file and attach that?

journalctl -b -t kernel > kernel.log

 

[philjans]

it seems there went something wrong with the output. can you redirect the command into a file and attach that?

journalctl -b -t kernel > kernel.log

Hi dcsapak, sorry for the delay: I am trying to figure out how to transfer the kernel.log files to this forum...

 

[philjans]

Ok got it. WinSCP worked.
Here is the log

 

[dcsapak]

ok the start of the kernel log seems to have rotated out since the last boot...

but i see in the logs some oom-killer invocations, how much memory do you have assigned?

we can also go try this another way (if you do not want to reboot)\

please post the output of:

dmidecode

 

[philjans]

Here it is:

 

[dcsapak]

ok all indidcates that that machine has that amount of memory... are you sure that the blade you showed in the first post correspond with the pve node?

 

[philjans]

ok all indidcates that that machine has that amount of memory... are you sure that the blade you showed in the first post correspond with the pve node?

I pretty sure it is but the admin who put it in place left: how can I confirm this?

 

[dcsapak]

I pretty sure it is but the admin who put it in place left: how can I confirm this?

for example the dmidecode returns the serial number (which should also be in the blade management)

also maybe the blade management interface is just wrong? (i do not know it and its behaviour)

in the worst case, turn off the blade, and look physically inside to see the dimms

 

[dcsapak]

oh, and to answer the second question in your post

Also, I read something that I am affraid would apply here: "I am using XenServer hypervisor and all these 4 servers are in a pool. I heard that, all servers in the pool should be in identical configuration by cpu, memory and all hardware configuraiton."
Is this applies to Proxmox also?

no, the servers do not have to have identical hardware configuration (though it helps to keep them similar)
only live migration can make problems between different cpu generations(e.g. sandy bridge <-> skylake)
best to keep the cpu type of the vms to kvm64 in that case

also between vendors (intel <-> amd) it is not guaranteed to work

 

[philjans]

Finally, like you said, opened it up and visually looked at it!

Proxmox was right with 56gb of ram and the iDrac of the dell blade server inventory was wrong.
I don't understand why the blade itself wouldn't report the right amount of ram but the OS would!??
I had like 6x4gb and 4x8gb
There was 2 slots free so I added 2x4gb and now I have 64...
thansk you