Do you periodically vmotion off and reboot esxi5 hosts?

[kjstech]

Hello,

We updated our ESXi 5.0 hosts back in February.  Recently as of last week two ESXi 5.0 hosts PSOD with NOT_IMPLEMENTED which hung up live VM's of course.  VMWare said there is no fix in our current build ESXi 5.0 patch 11.

They suggested upgrading to ESXi 5.1, 5.5 or 6.0.  Well I tried 5.1 upgrade over a year ago and it was unsucessfull at upgrading the vcenter server.  Could never get any certificates right, services had trouble starting, mostly SRM and VUM.  Oh well so we rolled it back to 5.0 and kept with it.  I don't want 5.5 or 6.0 because they have the insecure flash based web client.  We'll wait until VMWare goes HTML5 or better yet, returns to an updated Windows based thick client.  If they don't, we'll switch to Microsoft HyperV (we already paid for 2012 R2 Datacenter licences for all our ESXi hosts).

Anyway because 5.0 is the golden version for us that just works, and the uptime is well over 100+ days, do you think scheduled monthly reboots of hosts are worth it?  Do you do scheduled ESXi host reboots?  With vMotion it shouldn't be a big deal, but just wondered what some of your uptime records were.

[rcporto]

Normally I reboot vSphere ESXi host only when required for some patch/upgrade... if it is not required I leave my host running without reboot them, I have some hosts at version 4.1 running for 1.058 days, version 5.1 running for 490 days and 5.5 running for 187 days, and everything works like expected.

About upgrade to newer versions of ESXi/vCenter, I strictly recommend you do, about the vSphere Web Client, you can continue using the vSphere Client (thick version) with the whole full feature set of the version 5.0, you will only missing the new features added at version 5.1/5.5/6.0 that is only available on vSphere Web Client.

[aragab]

You should have a maintenance window for each single server in your environment every 90 days.

You should check the newly released driver/firmware levels for the servers's hardware components "BIOS. HBAs, RAID controllers and NICs" and apply the latest ones supported by your ESXi version; As you know with every driver and firmware level come a set of fixes to known issues.

[manmohanbisht]

Very rarely,reboot only if you have a specific reasons to restart hosts like  patch, hardware maintenance, driver/ firmware upgrade etc.

[Bleeder]

Short answer: Yes.

Even though there may not be any ESXi updates for months, there are usually other updates such as ESXi drivers, system firmware, etc from other sources.  These days, a good amount of those updates are usually marked as critical or security-related, so I end up being forced to do host maintenance/reboots at least monthly in order to stay in compliance with various audits (PCI DSS, government mandates, etc).

[Madmax01]

Hi theire,

in earlier Times of "ESX"  i planned more often to reboot the Hosts.

But to be honest now since "ESXI" i just do the Reboots once making Maintenance for Patching/Drivers/FW/Bios Activities.

The Versions  are getting more and more stable per Default, so mostly no Reason to Punch Periodicly Reboots. That's more modern :smileywink:.

Re Upgrade. So i just did red "certificate" problems.  i remember from earlier times i had to recreate all certs to an higher Security.  So i created for sha-256 with 2048 RSA. i just didn't had that much certificates necessary to change.

So if you still have "MD5" onces > then Vmware Update of vCenter don't let you through :smileywink:.       and for Esxi i would recommend to re/install fresh. No Upgrades.

If you're environment is supported for 5.5/6.0, then i would remind for planning to get onto minimum 5.5  ones. As much has changed.

Best regards

Max

[Kahonu84]

Aloha,

We only reboot as dictated by upgrades. As far as version goes, I would try upgrading to 5.5U2. You can

continue to use the stable awesome thick client with no problems.

Mahalo,

           Billl

[kjstech]

Thanks guys for all your posts.  I think we should put on the calendar a regular scheduled vmware updates manager check and reboot, rather than "do it when I feel like it".  If its a regimented schedule then we can be sure our uptime is consistent across the different hosts.

I don't want to go so long that we get a PSOD again.  I will also do some reading on upgrading to 5.5 U2.  Its good to hear you can use the thick client still. I may want to setup a test environment with some old servers first.