Root cause of: RE: How'd this for a bad day? AKA bad me
So, the root cause: ESX 3.5 OS was installed onto SAN volume that contained my VM's. The install of that OS (effectively) removes pointers that VM's need when they boot up. Best practice is to disconnect the SAN links when installing this version of the OS so this doesn't happen. In fact our SE did this but apparently didn't disconnect one far enough. If we had left the VM's running we could have used a VM converter to move them to a different storage location. ESX 4.0 doesn't allow this activity. Our SE feels really about out the work he created for me - personally I'm just really happy he's a stand up guy and explained what happened. You do this stuff long enough and something like this eventually happens - it's called experience. Dave From: Andrew S. Baker [mailto:asbz...@gmail.com] Sent: Friday, October 08, 2010 9:36 AM To: NT System Admin Issues Subject: Re: How'd this for a bad day? AKA bad me I've said it before, but I will say it again. In a highly virtualized, heavily consolidated world, we need more planning, more thinking and more time for effective execution. Cutting corners will become more and more painful, and will bite more and more organizations. Hopefully, enough near misses will teach enough entities to do the right thing. That's just my optimism speaking, however. It will be incumbent on each technology professional to advocate or fight for the right solutions, or have an excellent exit strategy planned out. :) ASB (My XeeSM Profile)http://XeeSM.com/AndrewBaker Exploiting Technology for Business Advantage... On Fri, Oct 8, 2010 at 11:27 AM, Raper, Jonathan - Eagle jra...@eaglemds.commailto:jra...@eaglemds.com wrote: +1 from here as well. A vCenter reboot should not require a host reboot. If it did, that would (IMHO) be a huge problem in the design and purpose behind VMware. Talk to VMware. If your maintenance is not current, get current. On a related note, YESTERDAY, one of our storage groups on our SAN ran out of space (fortunately I'm not in or over the group responsible for that anymore!), and thus took down a number of systems, all part of our core electronic medical record system, eClinicalWorks, all virtual... We were without that app for more than 6 hours, and are still dealing with database replication issues today as a result TGIF! Jonathan L. Raper, A+, MCSA, MCSE Technology Coordinator Eagle Physicians Associates, PA jra...@eaglemds.com www.eaglemds.com From: Jonathan Link [mailto:jonathan.l...@gmail.commailto:jonathan.l...@gmail.com] Sent: Friday, October 08, 2010 9:40 AM To: NT System Admin Issues Subject: Re: How'd this for a bad day? AKA bad me +1 I'm just getting caught up on emails this morning. vCenter reboot shouldn't necessitate a reboot of a host server. On Fri, Oct 8, 2010 at 9:34 AM, Jeff Bunting bunting.j...@gmail.commailto:bunting.j...@gmail.com wrote: Why do you need to power down VMs to reboot vCenter? vCenter might be the problem with the missing VMs. VMWare support might be able to help you with those. Jeff On Fri, Oct 8, 2010 at 5:51 AM, David Lum david@nwea.orgmailto:david@nwea.org wrote: I have 7 production systems running on 3 different ESX boxes in an ESX cluster, and 2 different logical SAN volumes (sorry am not SAN savvy, I just know I have two different SAN volumes to choose from when making a VM). Today, a SAN blows up and takes out half - our SharePoint server (heavily used), a Terminal Server , and an internal occasionally-used web server (Namescape rDirectory). Then somehow, when I was told to power down the other 4 VM's so our VMWare guy could reboot a vCenter server, 3 of the 4 remaining VM's decided to go AWOL (a combination of missing and disconnected). That took out my other two Terminal Servers and another lightly used internal web server. Did I mention I don't have the normal backups for these things because ...well...I'm an idiot and didn't confirm our backup guy installed backup software on these servers as I stood them up (process error on my part since I should confirm it's on there). None of these store data - they all talk to a backend SQL and the Terminal Servers are used to run apps that are slow if they run the same apps over VPN. SharePoint we got back quick because we do have a staging equivalent of it, so it was repoint to a config and content DB, DNS change, and done. I do have copious notes on how I built the others and can rebuild from scratch easily enough (I just finished the three TS boxes), but dude...six servers at once? The most frustrating part was discovering that the 4 systems that had been powered off could have been migrated before power off and there would have been no issue with them - the power down nuked 'em. Oh, and the lone surviving server - the PGP Universal Server that manages the encrypted machines. (Yes, the PGP machines will still boot w/out the server up, but still, I've been on
Re: Root cause of: RE: How'd this for a bad day? AKA bad me
Experience may not be the best teacher, but it is the most expensive one... On Fri, Oct 8, 2010 at 13:34, David Lum david@nwea.org wrote: So, the root cause: ESX 3.5 OS was installed onto SAN volume that contained my VM’s. The install of that OS (effectively) removes pointers that VM’s need when they boot up. Best practice is to disconnect the SAN links when installing this version of the OS so this doesn’t happen. In fact our SE did this but apparently didn’t disconnect one far enough. If we had left the VM’s running we could have used a VM converter to move them to a different storage location. ESX 4.0 doesn’t allow this activity. Our SE feels really about out the work he created for me – personally I’m just really happy he’s a stand up guy and explained what happened. You do this stuff long enough and something like this eventually happens – it’s called “experience”. Dave From: Andrew S. Baker [mailto:asbz...@gmail.com] Sent: Friday, October 08, 2010 9:36 AM To: NT System Admin Issues Subject: Re: How'd this for a bad day? AKA bad me I've said it before, but I will say it again. In a highly virtualized, heavily consolidated world, we need more planning, more thinking and more time for effective execution. Cutting corners will become more and more painful, and will bite more and more organizations. Hopefully, enough near misses will teach enough entities to do the right thing. That's just my optimism speaking, however. It will be incumbent on each technology professional to advocate or fight for the right solutions, or have an excellent exit strategy planned out. :) ASB (My XeeSM Profile) Exploiting Technology for Business Advantage... On Fri, Oct 8, 2010 at 11:27 AM, Raper, Jonathan - Eagle jra...@eaglemds.com wrote: +1 from here as well. A vCenter reboot should not require a host reboot. If it did, that would (IMHO) be a huge problem in the design and purpose behind VMware. Talk to VMware. If your maintenance is not current, get current. On a related note, YESTERDAY, one of our storage groups on our SAN ran out of space (fortunately I’m not in or over the group responsible for that anymore!), and thus took down a number of systems, all part of our core electronic medical record system, eClinicalWorks, all virtual… We were without that app for more than 6 hours, and are still dealing with database replication issues today as a result…. TGIF! Jonathan L. Raper, A+, MCSA, MCSE Technology Coordinator Eagle Physicians Associates, PA jra...@eaglemds.com www.eaglemds.com From: Jonathan Link [mailto:jonathan.l...@gmail.com] Sent: Friday, October 08, 2010 9:40 AM To: NT System Admin Issues Subject: Re: How'd this for a bad day? AKA bad me +1 I'm just getting caught up on emails this morning. vCenter reboot shouldn't necessitate a reboot of a host server. On Fri, Oct 8, 2010 at 9:34 AM, Jeff Bunting bunting.j...@gmail.com wrote: Why do you need to power down VMs to reboot vCenter? vCenter might be the problem with the missing VMs. VMWare support might be able to help you with those. Jeff On Fri, Oct 8, 2010 at 5:51 AM, David Lum david@nwea.org wrote: I have 7 production systems running on 3 different ESX boxes in an ESX cluster, and 2 different logical SAN volumes (sorry am not SAN savvy, I just know I have two different SAN volumes to choose from when making a VM). Today, a SAN blows up and takes out half – our SharePoint server (heavily used), a Terminal Server , and an internal occasionally-used web server (Namescape rDirectory). Then somehow, when I was told to power down the other 4 VM’s so our VMWare guy could reboot a vCenter server, 3 of the 4 remaining VM’s decided to go AWOL (a combination of “missing” and “disconnected”). That took out my other two Terminal Servers and another lightly used internal web server. Did I mention I don’t have the normal backups for these things because …well…I’m an idiot and didn’t confirm our backup guy installed backup software on these servers as I stood them up (process error on my part since I should confirm it’s on there). None of these store data – they all talk to a backend SQL and the Terminal Servers are used to run apps that are slow if they run the same apps over VPN. SharePoint we got back quick because we do have a staging equivalent of it, so it was repoint to a config and content DB, DNS change, and done. I do have copious notes on how I built the others and can rebuild from scratch easily enough (I just finished the three TS boxes), but dude…six servers at once? The most frustrating part was discovering that the 4 systems that had been powered off could have been “migrated” before power off and there would have been no issue with them – the power down nuked ‘em. Oh, and the lone surviving server – the PGP Universal Server that manages the encrypted machines.