ESXi

[E_Schmidt]

Dear ladies and gentlemen,

in March 2017 we bought an ESXi Licence and thought we will have support with it. So we extented the "support" in 2018 for another 3 years. But as we know now, the "Subscription only" plan is only for updates, not for support. Damn ...

That's why I'm opening this thread. We are running 2 ESXi Hosts, both with Raid 5. I know, Raid 5 is not well known for it's speed, but yeah, read the next few lines.

We get a speed of about 20MB/s when copy files from one of our hosts or even at this host between datastores. Running on a ThinkSystem SR650. The ESXi is connected with fibre channel to one of our switches, no problem with other hosts. Like i said, even when I connect with SSH to the ESXi and try to duplicate an VMDK on file level, it's very slow. Can't measure the exact speed, but it's slow, trust me. The other host is faster, downloading a VMDK of it gives us a speed of about 100M/sec.

Is there any way to export something like a debug profile for you guys here on the forums? I think that would be helpful. We are running very critical tasks on this ESXi, so a reboot is not an option at the moment. We need to replicate our VMs to the 2nd Host before we can do this. But maybe there are some steps without reboot which we can try.

Thanks for any advice!

[a_p_]

Welcome to the Community,

issues like the one you describe are often caused by a RAID controller operating in "Write-Through" mode, rather than "Write-Back" mode.

In case the controller's cache is backed by a battery, check the battery status. A bad battery usually changes the RAID controller to operate in "Write-Through" mode to protect data in case of a power loss.

André

[E_Schmidt]

Hello and thanks for the nice welcome. I'm thinking about your answer, but does it make sense when downloading a VMDK with browser that the write-mode may be wrong? It should read and not write. Or am I wrong?

[a_p_]

Good point. You're right, reading from disk shouldn't be affected that much. Anyway, a quick look at the hardware's health will not hurt.

I assume that you've already checked the network ports/SFPs for e.g. CRC errors!?

What may contain hints on what may be causing the issue is the ESXi host's vmkernel.log.

André

[E_Schmidt]

Hi, oh yes, I totally forgot about the vmkernel.log file. I will attach it to this answer.

Switch seems to be fine, no errors belonging to the ESXi, no CRC errors. Using some Aruba-2930F-24G-4SFPP here.