Uptime recorded differently between 1.3 and 2.0

[Rob_Bohmann1]

I just noticed this today. Whenever I VMotion a guest it resets the uptime clock. This occurs whether I manually move a guest or if I put a host into maintenance mode.

I understand it is calculating the uptime for only the host that it is currently running on. It would be nice to be able to see how long it has been up across a cluster, like it seems to work in VC 1.3. Just posting this to verify that this is what others see and let everyone know about this change between versions.

[hicksj]

Soft reboot one of your VM's and what does the uptime show?

For me, if it were 15 days pre-reboot, it remains at 15 days post-reboot.

Uptime is monitoring how long that VM has been powered up on that particular host. I guess if you want "real" uptime, use something on the OS. (snmp, uptime util, etc)

[Rob_Bohmann1]

Yes i see the same thing, when I reboot an OS. The uptime does not change. I saw this today when I was VMotioning guests in a couple of different clusters. After a couple of hours it showed the uptime as a couple of hours when previously it was several days.

[Illaire]

I just noticed this today. Whenever I VMotion a

guest it resets the uptime clock. This occurs whether

I manually move a guest or if I put a host into

maintenance mode.

Yes. I've seen the same thing. At first, I had just looked to check if our infrastructure was OK and nearly had a panic attack seeing our critical servers with an uptime of a few minutes.

It turns out that those servers were just VMotioned and Virtual Center 2 reports uptime from the starting time on a specific server. This is confusing sometimes.

I didn't found a way to change that in Virtual Center 2. That's just the way it works.

[Rob_Bohmann1]

THanks for checking this and the feedback, just wanted to be sure this was not something just seen in our environment.

[gogogo5]

Yes, it also happens here too (running ESX 3.01 and VC 2.01).

Vmotioning causes the uptime value to reset and you'll see the Guest Memory % value jump to around 75% about 40 seconds after the vmotion.

[jasonboche · Accepted solution]

I just noticed this today. Whenever I VMotion a

guest it resets the uptime clock. This occurs whether

I manually move a guest or if I put a host into

maintenance mode.

I understand it is calculating the uptime for only

the host that it is currently running on. It would

be nice to be able to see how long it has been up

across a cluster, like it seems to work in VC 1.3.

Just posting this to verify that this is what

others see and let everyone know about this change

between versions.

Noticed this tonight and I hate it.

Uptime should only be reset by a VM hardware power off operation (which can be performed by a power off or a guest shutdown which in turn powers off the VM).

I'm submitting the feature/bug request on this one.

Jas

[Rob_Bohmann1]

Since you are submitting the change request, I'll give you some thank you points!

Thanks!

[jasonboche]

Thanks! Let's hope they get this bug fixed soon. Should be an easy one to nail down.

Jas

[hicksj]

I don't know that it will be so easy. VMware can guarantee knowledge of how long a VM has been active on its host, but cannot inspect true VM guest OS uptime 100%. VMware tools would have to be used to report OS uptime, and not all VMs may be running this.

I guess they might be able to enhance the VMotion process so it passes the previous uptime along...

An then, what happens when you reset the guest OS? Do you want to see this counter drop to 0, or remain where it is? After some more thought on this, I kinda dig knowing how long a VM has been active on a specific host... especially when DRS is part of the equation.

Regards,

J

[Rob_Bohmann1]

Yeah depending on the circumstances, you might want to know time on the host or total time since power on (or since last boot).

The problem is while we can track all sorts of info, you need to make choices to limit to what is truly useful for most of the people most of the time. Otherwise this VC database is going to grow and grow and grow, and performance and reliability become more at risk or the resources you have to throw at it increase.

I can make arguments for all 3 measurements, but if you want to call it uptime, then it should be from last power on from a virtual center perspective, from an OS Perspective it is last boot, and from an ESX perspective is it since last VMotion. My personal vote is power on since I am looking at it from within VC.

my$.02

[jasonboche]

Agreed that various people will find a need or interest in various column headings/statistics, but in this case, the column label clearly doesn't match the value. If the value is going to reflect "since last power on or VMotion", then that's what the column description should be. That's not the same as uptime. I think we all know what the industry definition of uptime is and that term shouldn't be futzed with. It's confusing enough they went and used the name clusters in VI3. :smileyshocked:

[hicksj]

I have a solution to the problem.

Right click on the headers and deselect "Uptime"

[jasonboche]

no

[jasonboche]

It looks like a soft reboot of the VM inside of the OS also resets the uptime statistic in VirtualCenter. Likely a function of VMware Tools that triggers this.

I would bet that a VM which has no VMware Tools installed that is soft rebooted inside of the OS (ie. We're not talking about any VMware console power operation here) would not reset the uptime statistic.

This particular one I could see the community more on the fence about as there may be a slight variation of the definition of uptime here. Does uptime translate into how long the VM has been powered on from a virtual infrastructure standpoint or does uptime translate into how long the OS inside the VM has been up which is tracked by the VMware Tools heartbeat?

The VMware administrator cares about the VM power on.

The Windows or Linux adminsitrator cares about uptime of the OS.

I think the intended use of the Virtual Infrastructure Client is for the VMware administrator and thus should reflect virtual infrastructure information, therefore, my personal preference is that the "uptime" statistic in VirtualCenter tracks power operations of the VM irregardless of any reboot sequences of the OS inside the VM.

Jas

[wilson94t]

Agree, there are all kinds of OS tools to determine OS uptime. Does anyone have a work around for this issue? I'd like to be able to do the query to find out how long the virtual hardware has been available on request - giving me the same information as the Virtual Center 1.x uptime.