#6 Issue with nmon and RHEL7 autofs

[Yann Renard]

Hello,

In my RHEL6 system I used to run this command to capture data from a system : nmon -m /tmp_nmon -f -s 90 -c 960
In my RHEL7 system when I use the same command all the fs referenced by autofs are immediatly mounted. I have the same exact autofs configuration.

I think it's a bug. Can you help me with that ?

myhost # df
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/mapper/rootvg-rootvg 5029504 3101060 1649920 66% /
devtmpfs 5026704 0 5026704 0% /dev
tmpfs 5037652 25216 5012436 1% /run
tmpfs 5037652 0 5037652 0% /sys/fs/cgroup
/dev/mapper/rootvg-tmp 999320 2852 927656 1% /tmp
/dev/mapper/rootvg-var 3997376 1009956 2761324 27% /var

myhost # ./nmon_x86_rhel72 -m /tmp_nmon -f -s 90 -c 960
[ ~ ]
root@dal-b-cl04 # df
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/mapper/rootvg-rootvg 5029504 3101092 1649888 66% /
devtmpfs 5026704 0 5026704 0% /dev
tmpfs 5037652 25216 5012436 1% /run
tmpfs 5037652 0 5037652 0% /sys/fs/cgroup
/dev/mapper/rootvg-tmp 999320 2852 927656 1% /tmp
/dev/mapper/rootvg-var 3997376 1009956 2761324 27% /var
a.b.c.d/nfsvolume 524288000 411290360 112997640 79% /test

myhost # ./nmon_x86_rhel72 --version
./nmon_x86_rhel72: invalid option -- '-'
Hint for nmon_x86_rhel72 version 16d
Full Help Info : nmon_x86_rhel72 -h

    On-screen Stats: nmon_x86_rhel72
    Data Collection: nmon_x86_rhel72 -f [-s <seconds>] [-c <count>] [-t|-T]
    Capacity Plan  : nmon_x86_rhel72 -x

Interactive-Mode:
Read the Welcome screen & at any time type: "h" for more help
Type "q" to exit nmon

For Data-Collect-Mode
-f Must be the first option on the line (switches off interactive mode)
Saves data to a CSV Spreadsheet format .nmon file in then local directory
Note: -f sets a defaults -s300 -c288 which you can then modify
Further Data Collection Options:
-s <seconds> time between data snapshots
-c <count> of snapshots before exiting
-t Includes Top Processes stats (-T also collects command arguments)
-x Capacity Planning=15 min snapshots for 1 day. (nmon -ft -s 900 -c 96)
---- End of Hints

Regards.

[Nigel Griffiths]

Hi,
I have to assume you are running the same nmon version 16d on both RHEL6 and 7 plus you are running in 32 bit.

The filesystem gathering code in nmon has been stable and working for many years and it uses classic UNIX code to perform this operation.
getmntent() get mount entry to list the mount points and the fstatfs() to get the filesystem stats.

I don't follow that this is a nmon bug at all.

So if this operates differently on RHEL7 then RHEL6 then the bug is in RHEL6 or RHEL7.

You have a logic bomd here that you might not be aware of.
Don't expect nmon to cater for filesytsems that come and go during the life of the data collection.

The vast bulk of the tools that use nmon collected data can't cope with sudden mid collection additional columns of data. That would require nmon to dynamically deal with the change by tracking before and after stats names and make the nmon output much move complex. This would result in nmon taking lots more CPU cycles - which no one wants.

All the down stream tools would have to change too.

So the options (it seems) are to ignore autofs filesystems completely like RHEL6 and there is unexpained I/O to mystery filesystems or like RHEL7 mount the autofs filesystems as nmon starts and asks for mount point details so we capture data for all of them as an when filesystems get updated. I don't know of a way to discover the filesystem in RHEL7 that does not mount them - that is a RHEL7 glibc library feature.

Either way I don't see this as an nmon bug. If you insist on the RHEL7 being consistent with RHEL6 then you need to raise a RHEL7 bug report.

Cheers, Nigel Griffiths