When building a new lab environment I ran into problems with VUM sometime failing to scan the host. After searching through the logs, it showed that the ramdisk was full:
2015-08-03T10:04:16.746Z cpu0:119094)WARNING: VisorFSRam: 351: Cannot extend visorfs file /usr/share/vua/libcrypto.so.1.0.1 because its ramdisk (root) is full.
2015-08-03T10:04:16.758Z cpu1:119094)WARNING: VisorFSRam: 351: Cannot extend visorfs file /usr/share/vua/libssl.so.1.0.1 because its ramdisk (root) is full.
2015-08-03T10:04:16.769Z cpu0:119094)WARNING: VisorFSRam: 351: Cannot extend visorfs file /usr/share/vua/vua because its ramdisk (root) is full.
After further cumbersome investigation I found the root cause: when the esx-splitter / kdriver is installed, over half the ramdisk is constantly used. When VUM now tries to scan a host there is either just enough free space to run (1M free left when VUM / vua is running) for it to succeed or not.
Before Install:
Ramdisk
Size
Used Available Use% Mounted on
root
32M
180K
31M 0% --
etc
28M
320K
27M 1% --
opt
32M
0B
32M 0% --
var
48M
620K
47M 1% --
tmp
256M
80K
255M 0% --
hostdstats
553M
1M
551M 0% --
After Install:
Ramdisk
Size
Used Available Use% Mounted on
root
32M
17M
14M 55% --
etc
28M
316K
27M 1% --
opt
32M
0B
32M 0% --
var
48M
984K
47M 2% --
tmp
256M
100K
255M 0% --
hostdstats
178M
1M
176M 0% --
This is true for all ESXi versions (5.1, 6.0), installation methods (direct to local disk, usb-stick + configured scratch location), splitter versions (4.2.0.m104, 4.3.0.c112) and ESXi-Images (Standard, IBM/Lenovo customized).
VUM Failing
Ramdisk
Size
Used Available Use% Mounted on
root
32M
20M
11M 64% --
etc
28M
300K
27M 1% --
opt
32M
0B
32M 0% --
var
48M
480K
47M 0% --
tmp
256M
20K
255M 0% --
hostdstats
178M
1M
176M 0% --
The additional 3M disappear when the host is rebooted.
Any hints how to find out which files allocate that extra 3MB are appreciated.
To be honest, I can't remember if this caused problems in the old lab environment.
But with the extensive effects that a full ramdisk can have and their possible impacts in a production environment, I really wonder if this is expected behavior? Anf if yes: can this maybe changed with some kind of tweak-parameter?
P.S.: What is the intended use of the offline-bundle provided with the 4.3 deployment kit? In the provided state it is pretty useless: It shows not applicable for 6.0 (because this version is missing from meta-data) and even with tuning the meta-data the splitter can't be installed via VUM (because the bundle does not have a signature).
For the signature issue: shouldn't be a problem for EMC to make a compatible and working ('just works') offline-bundle for VMware (Update Manager) . SCP/SSH as tools to install the splitter are no rocket-science, but not the best way when the splitter should be installed on a considerable amount of hosts.
forshr
2 Intern
•
1126 Posts
1322
0
Posted August 10th, 2015 07:00
Hi Daniel,
The ramdisk issue will be resolved in 4.3 SP1. As for the signature issue, we are looking into it and I will come back to you.
Regards,
Rich Forshaw
RecoverPoint Corporate Systems Engineering