VMware Cloud Community
UCL
Enthusiast
Enthusiast

Upgrading ESXi 6.5U2 to ESXi 6.7U1. Completes Upgrade. Insufficient ramdisk space but ScratchConfig.CurrentScratchLocation is configured and persistent.

Hello.

Using HPE 6.7U1 image "VMware-ESXi-6.7.0-Update1-10302608-HPE-Gen9plus-670.U1.10.3.5.12-Oct2018.iso" on a ESXi 6.5U2 Host with a "HPE-ESXi-6.5.0-Update2-iso-Gen9plus-650.U2.10.3.5.5" the Update Manager reports it as Non-Compliant which correct.

But after the upgrade I have a 6.7 Host which reports as "Incompatible" due to "Cannot create a ramdisk of size 359MB to store the upgrade image. Check if the host has sufficient memory."

But I have set ScratchConfig.CurrentScratchLocation and it persists after reboots.

When I reboot the host after 6.7 the host revert to 6.5.

A vdf -h shows:

Ramdisk                   Size      Used Available Use% Mounted on

root                       32M        2M       29M   7% --

etc                        28M      364K       27M   1% --

opt                        32M      532K       31M   1% --

var                        48M      632K       47M   1% --

tmp                       256M      728K      255M   0% --

iofilters                  32M        0B       32M   0% --

shm                      1024M        0B     1024M   0% --

hostdstats               1303M       12M     1290M   0% --

snmptraps                   1M        0B        1M   0% --

Is it actually the SD-Card which lacks space or what ?

/Peter

Reply
0 Kudos
21 Replies
ThompsG
Virtuoso
Virtuoso

Hi Peter,

Coming in from left field but do you happen to have "Lockdown mode" enabled on this host?

Kind regards.

Reply
0 Kudos
pragg12
Hot Shot
Hot Shot

Hi,

Also check if the root account on ESXi still has "administrator" access.

Check /etc/vmware/hostd/authorization.xml. For root, the ACEDataRoleId value should be set to '-1', which means Administrator.

<ACEData id="10">

    <ACEDataEntity>ha-folder-root</ACEDataEntity>

    <ACEDataId>10</ACEDataId>

    <ACEDataIsGroup>false</ACEDataIsGroup>

    <ACEDataPropagate>true</ACEDataPropagate>

    <ACEDataRoleId>-1</ACEDataRoleId>

    <ACEDataUser>root</ACEDataUser>

  </ACEData>

Source: https://communities.vmware.com/thread/494782

Edit: Formatting.

Consider marking this response as "Correct" or "Helpful" if you think my response helped you in any way.
Reply
0 Kudos
UCL
Enthusiast
Enthusiast

Hello ThompsG

Lockdown mode on that Host is disable.

/Peter

Reply
0 Kudos
UCL
Enthusiast
Enthusiast

Hello pragg12.

Those settings are identical on our Host.

/Peter

Reply
0 Kudos
pragg12
Hot Shot
Hot Shot

Hi UCL

Is your issue resolved or still facing ? Let me know so I can further assist accordingly.

Consider marking this response as "Correct" or "Helpful" if you think my response helped you in any way.
Reply
0 Kudos
UCL
Enthusiast
Enthusiast

Hello.

VMware Technical Support is working on it.

/Peter

Reply
0 Kudos
pragg12
Hot Shot
Hot Shot

Hi,

Good to know. Once solved, let us know the cause and resolution as provided by VMware support.

Consider marking this response as "Correct" or "Helpful" if you think my response helped you in any way.
Reply
0 Kudos
UCL
Enthusiast
Enthusiast

Personnally I'm currently investigating whether this devListStabilityCount=10 needs to be added to /bootbank/boot.cfg

I saw that before upgrade  bootbank -> /vmfs/volumes/0c517feb-0fb973e9-b89d-76f429f6fd45 and after bootbank -> /tmp.

After reboot it reverted back to bootbank -> /vmfs/volumes/0c517feb-0fb973e9-b89d-76f429f6fd45.

That led to this article: VMware Knowledge Base

But I'm not sure that applies to upgrades to 6.7U1 so I'm testing if adding that  devListStabilityCount=10  makes a difference.

/Peter

Reply
0 Kudos
UCL
Enthusiast
Enthusiast

Hmm nope. Same same.

I didn't update any patches though but only added that option.

/Peter

Reply
0 Kudos
UCL
Enthusiast
Enthusiast

2019-03-18T11:29:46Z esxupdate: 68546: BootBankInstaller.pyc: INFO: Unrecognized value "devListStabilityCount=10" in boot.cfg

Reply
0 Kudos
pragg12
Hot Shot
Hot Shot

Can you share an update on your issue status ?

Consider marking this response as "Correct" or "Helpful" if you think my response helped you in any way.
Reply
0 Kudos
UCL
Enthusiast
Enthusiast

Yes.

There are two LUNs on that SC Card which the ESXi Host boots from and no utility to delete the last LUN. You can delete the partitions on that last LUN but not the LUN itself.

The main problem is that the upgrade does not see the first LUN and upgrade is performed to /tmp instead of the Bootbank.

So the upgrade succeeds until you reboot the host. Then it boots up as ESXi 6.5U2 instead of 6.7U1

Neither upgrade through Update Manager nor by mounting the ISO file on the Host to upgrade makes any difference. Only a clean install does.

I installed on the two troublesome Host and they work now.

I have 6 other hosts with only one LUN each on their SD Cards and those were upgraded without any insident.

/Peter

Reply
0 Kudos
UCL
Enthusiast
Enthusiast

Turned out to be a HPE Embedded User Partition. Disable it as a Boot source and delete it in the server BIOS. That solves the problem.

Enabling or disabling the Embedded User Partition

https://support.hpe.com/hpsc/doc/public/display?docId=c04398276  page 60 Delete Boot Option.

/Peter

1 person fandt dette nyttigt

pragg12
Hot Shot
Hot Shot

Thanks for sharing the root cause and details. It will definitely be helpful to the community folks.

On another note, HPE being HPE. ¯\_(ツ)_/¯

Edit: Formatting

Consider marking this response as "Correct" or "Helpful" if you think my response helped you in any way.
Reply
0 Kudos
JeffST
Enthusiast
Enthusiast

i got a simlar problem on a dl360 gen10. upgrading from 6.5 to 6.7 update 2 reverted back to 6.5 on second reboot after the upgrade. bootbank got mounted to /tmp, all nasty.

Gen10 with ilo 5 do not have embedded partitions etc which can be enabled/disabled. So that was not the problem/fix here.

troubleshooted for quite some time and found out that after upgrading to 6.7 update 2 , the vmkusb driver was not loaded. esxcfg-scsidevs -a did not output an entry for the mirrored usb boot device as it did in 6.5. so i suspected the problem was related to vmkusb driver in 6.7 combined with the 8GB mirrored microsd usb adapter. No further root cause known to me at the moment.

more info about the vmkusb driver : VMware Knowledge Base

what i did to perform the upgrade succesfully eventually:

i rebooted the 6.5 host, hitting shift+O on the bootscreen of ESXi

i added preferVmklinux=TRUE as a boot parameter.

the host wil boot using the vmklinux drivers for the usb storage instead of the vmkusb. You will see the output of esxcfg-scsidevs -a change to a vmhba32 entry. the driver used for the bootdevice is now called "usb-storage" instead of "vmkusb".

i disabled the vmkusb driver with

esxcli system module set -m=vmkusb -e=FALSE

i performed the upgrade to 6.7 (in my case using the cli and a zip file as a depot, but should work with VUM as well i think) and rebooted. bootbank and altbootbank mounted correctly and everything seemed ok, also performed a second reboot. looking good.

checked the kernelparameter using:

esxcli system settings kernel list | grep prefer

check to see if preferVmklinux is set to enabled. it was in my case, if not use:

esxcli system settings kernel set -s preferVmklinux -v TRUE

The disadvantage with this solution is you lose Quickboot as a feature. that's really disappointing.

I also tried to revert to vmkusb after the upgrade, just curious, but that resulted in not loading vmkusb again and bootbank mounting to /tmp etc. no good.

So stuck with legacy usb drivers it seems and no quickboot Smiley Sad

i'm done with the sd-bootcards and usb mirrored boot solutions  for sure! Smiley Happy

Better pay up, and stay up!

Reply
0 Kudos
pragg12
Hot Shot
Hot Shot

Hi,

Thanks for sharing this!

Have you got this checked with VMware support if there is a workaround possible so you don't lose quickboot ?

Consider marking this response as "Correct" or "Helpful" if you think my response helped you in any way.
Reply
0 Kudos
JeffST
Enthusiast
Enthusiast

i have not contacted support for this yet. Requirements for quickboot are pretty clear, it cant work with the legacy drivers enabled. The strategy for the legacy drivers is clear as well. I dont have high hopes for any engagement with support. I might do it later, but im hesitant because i suspect it will take too much of my working with support, collecting logs just to get to the point i'm right now.

Reply
0 Kudos
Mallik7
Enthusiast
Enthusiast

I found something on web similar to your question, see if this useful to you - https://www.nuttycloud.com/cannot-create-a-ramdisk-of-size-387mb-to-store-the-upgrade-image-check-if...

I found the same kind of error when I'm trying to upgrade from 6.0 to 6.5 a month ago.

Reply
0 Kudos
aadhish95
Contributor
Contributor

tmp folder 100 % usage is the reason for Ram Disk full , You need to Move a file from tmp folder to your local , I will share commands

Reply
0 Kudos