VMware Cloud Community
kwg66
Hot Shot
Hot Shot
Jump to solution

scratch partition, syslog, and VMKdump clarification needed

I need some simple clarification from VMware:

The scratch partition is automatically created during the installation if the space is there, but if not it is created in RAMDISK and won't be persistent..

OK, if this is true, how come even with plenty of space on my SD Cards housing my installation I am told by VMware support that its not persistent?

Finally, if global syslog is configured, is the scratch partition needed at all?   Are they different in what they do, or do the have the exact same logs? 

Also, what about VMKDump... is this now defaulted to a datastore?  and can anyone speak about the value of setting up the network dump?  

1 Solution

Accepted Solutions
Finikiez
Champion
Champion
Jump to solution

>if scratch is on SD card it goes to ramdisk which is not persistent  -- right

>if global syslog target is set up the logs ‘prior to reboot‘ should be available there correct -- logs will be there prior reboot and after reboot as well Smiley Happy

if you configured syslog.global.logdir - logs always will be there. However there are some scenarios. For example you store ESXi logs on a VMFS datastore connected through FC to ESXi host. If you lose access to this datastore, there will be no new messages there.

Here is a good KB about configuring syslog https://kb.vmware.com/kb/2003322

View solution in original post

0 Kudos
4 Replies
Finikiez
Champion
Champion
Jump to solution

Hello!

1. SD is a flash device and to avoid SD card failure because of intensive write operations, scratch partion will never be created on it. It doesn't matter if you have free space on it.

2. Syslog directory is just a symlink for a scratch partition

for example

# ls -lah /

...

lrwxrwxrwx    1 root     root          49 Sep 28 08:46 scratch -> /vmfs/volumes/53302444-5f979f4e-c9b3-002128a2b0dc

3. Network dump collector is an additional place where to store crash dump in case of PSOD.

There is one thing with using network dump collector to think about - when host crashes into PSOD and writing dump to a network dump collector it's available for a ping and so HA doesn't restart VMs until host finishes writing a dump.

kwg66
Hot Shot
Hot Shot
Jump to solution

thanks for the thorough reply.  Final question around this topic-

if scratch is on SD card it goes to ramdisk which is not persistent, but if global syslog target is set up the logs ‘prior to reboot‘ should be available there correct?  

0 Kudos
Finikiez
Champion
Champion
Jump to solution

>if scratch is on SD card it goes to ramdisk which is not persistent  -- right

>if global syslog target is set up the logs ‘prior to reboot‘ should be available there correct -- logs will be there prior reboot and after reboot as well Smiley Happy

if you configured syslog.global.logdir - logs always will be there. However there are some scenarios. For example you store ESXi logs on a VMFS datastore connected through FC to ESXi host. If you lose access to this datastore, there will be no new messages there.

Here is a good KB about configuring syslog https://kb.vmware.com/kb/2003322

0 Kudos
kwg66
Hot Shot
Hot Shot
Jump to solution

Thanks - beer on me if you come to Boston!