VMware Cloud Community
silicoon
Contributor
Contributor
Jump to solution

ESX partitioning - vmkdumps overfilling into root?

hi guys

an esx consultant was brought into work the other day and said that the ESX partitioning on our servers was bad and caused the root to overfill because the ESX had a ridiculous number of vmkcore dumps?

now over looking the underlying problem of why its dumping - i was under the impression that by default the vmkcore partitioning was setup so that vmkdumps couldnt overfill into root? the partitioning that was setup was ESX defaults with extra space added to swap / var / etc etc

I do know that you should create separate partitions for var tmp opt home or whatever so that it doesnt overfill into root but i thought vmkcore was preconfigured in ESX to stop this?

is this guy talking crap? or am i tripping out

thanks in advance

Reply
0 Kudos
1 Solution

Accepted Solutions
JacquesCronje
Enthusiast
Enthusiast
Jump to solution

vmkdump is the host crash dump - last time I checked it dumped around 100MB for inclusion in the vm-support tarball. If you are running ESX3.x you have nothing to worry about in terms of the RH service console partitioning - that was something we paid a lot of attention to in the ESX2.x days. Sounds like scaremongering to me.

View solution in original post

Reply
0 Kudos
9 Replies
kjb007
Immortal
Immortal
Jump to solution

Typically, vmkcore is its own filesystem, and it will not overfill into /. That being said, it could be that there were different core files of vm's that were causing space problems. Do you know if / actually was filling? And filling with what files and in which location?

-KjB

vExpert/VCP/VCAP vmwise.com / @vmwise -KjB
silicoon
Contributor
Contributor
Jump to solution

the consultant said that vmkcore dumps were overfilling and corrupting some of the config files etc

This is a quote from him "esx01 had a full file system due to thousands of core dumps"

seems a bit weird to me that the default esx partitioning scheme would hold such a big flaw...

Reply
0 Kudos
XFoS
Contributor
Contributor
Jump to solution

By default, the partitioning set up during setup of ESX does create a vmkcore partition that's separate to the root partition. However, if the person who built the ESX hosts changed the default partitioning scheme and didn't include a vmkcore partition (this may be done by someone who understood Linux but not VMware, or by someone who understood neither and didn't recognise the importance of the vmkcore partition being separate).

JacquesCronje
Enthusiast
Enthusiast
Jump to solution

vmkdump is the host crash dump - last time I checked it dumped around 100MB for inclusion in the vm-support tarball. If you are running ESX3.x you have nothing to worry about in terms of the RH service console partitioning - that was something we paid a lot of attention to in the ESX2.x days. Sounds like scaremongering to me.

Reply
0 Kudos
JacquesCronje
Enthusiast
Enthusiast
Jump to solution

Also -

This is a quote from him "esx01 had a full file system due to thousands of core dumps"

This would mean your host has crashed 1000s of times - if that's true, you have a bigger problem than a poorly partitioned SC.

Reply
0 Kudos
XFoS
Contributor
Contributor
Jump to solution

"This is a quote from him "esx01 had a full file system due to thousands of core dumps""

Ask him to show you this - he should be providing evidence of everything he says. Also log into the service console and run 'df -h' as well - that will tell you exactly how the partitions are defined and how they're utilised.

Reply
0 Kudos
kjb007
Immortal
Immortal
Jump to solution

The core dumps could be from vmx process crashes as well, as well as other processes. As the others stated already, that kind of statement should merit some proof.

-KjB

vExpert/VCP/VCAP vmwise.com / @vmwise -KjB
Reply
0 Kudos
silicoon
Contributor
Contributor
Jump to solution

hi guys

i just found out that it was something else dumping to /var/core so it wasnt vmkcore dumps :S

another query - is there anyway that i can create a new partition that mounts /var without reinstalling esx... so that this doesnt happen again?

also is this partitioning scheme simular to the one that you guys would use?

Mount Point

Partition

Size

Description

/dev/sda (Primary)

/boot

ext3

250 MB

Change for additional space for upgrades

N/A

swap

1600 MB

Change for maximum service console swap size

/

ext3

5120 MB

Change for additional space in root

/dev/sda (Extended)

/var

ext3

4096 MB

Create partition to avoid overfilling root with log files

/tmp

ext3

1024 MB

Create partition to avoid overfilling root with temporary files

/opt

ext3

2048 MB

Create partition to avoid overfilling root with VMware HA log files

/home

ext3

1024 MB

Create partition to avoid overfilling root with agent / user files

vmkcore

100 MB

Pre-configured

Free Space

VMFS

Reply
0 Kudos
kjb007
Immortal
Immortal
Jump to solution

Yes. Create a new partition. Copy your data from /var to that new folder. Since there are runtime files stored under there, it may require a reboot, so the new files are correctly written to the new location.

-KjB

vExpert/VCP/VCAP vmwise.com / @vmwise -KjB
Reply
0 Kudos