Has a 7.x version been deployed on en VMware (production) architecture?

[ceesvl]

Hi Community,

As a user of version 6. We would like to know if anybody has deployed a 7.x version on a VMware cluster architecture (for production) and if there are known bugs or limitation in this kind of deployment. This is also call out to the Exasol experts if this version contains changes or limits that would impact a VM configuration.

Best regards,
Cees van Leuveren

[exa-Chris]

Hi Cees,
thanks for the question. I will check with the guys internally and ask the Xperts.
exa-Chris

[mwellbro]

Not yet, unfortunately - will let you know if we find something once we get around to it, but we´re currently in our last "pre 7" upgrade running on VMWare.

[exa-Lennart]

Hi @ceesvl,

on Thursday last week we successfully deployed an Exasol 7.0.1 on a prospects VMware infrastructure without any issues. This was a fresh installation - not an update. We did not encounter any issues. Testing will start this week so this is all I can say for now

[ceesvl]

Hi @exa-Lennart ,

Thanks for sharing. Is the prospect also planning to do some performance optimisation of the installation? If so we would love to share some ideas and/or findings.

[mwellbro]

Hi @ceesvl  , 

you got me at "performance optimisation" in combination with VMWare - do you have anything specific in mind ? We never ran "bare hardware" vs VMWare so I don´t know how much we "loose" to virtualization (if indeed anything of significance) or if there are certain "switches" like VMWare HardwareVersions that come into play, so I´d be interested in any findings you might happen upon 🙂

Cheers,
Malte

[ceesvl]

Hi All,

An update from my side.

It seems that VMWARE issue https://kb.vmware.com/s/article/2011861 is a big cause of performance loss for our installation. The command mentioned in the white paper  increases our performance by 60% and increases the max workload by 3 times before saturation occurs in a heavy load testing case.

After discussion with Exasol, it was decided that this is not a limitation of the core product and will not be fixed. Sadly we will have to make do with performing the noop scheduler command (root access is needed) on all nodes, every time we do a reboot.

On a side note, when my IT guys found out about this article they did a massive upgrade of almost all linux installations to the latest Ubuntu server release (that version is now VMWARE aware). This was a great savings on hardware usage and a performance boost to a lot of systems.

Best,

Cees