Recent changes to feature-requests

Monitoring voluntary unvoluntary ctx switch

Anonymous — Tue, 06 Dec 2011 17:05:01 -0000

Can be useful for tracking contention on thread to have this metrics
regards
i ve tried but i an not a perl guy ...
regards

rpm relocation support

Anonymous — Thu, 16 Jun 2011 08:44:49 -0000

It would be handy if collectl could be installed from rpm with relocation possibility:
rpm --prefix,
rpm --relocate x=y

Not possible in latest version [collectl-3.5.1]

Collection of NUMA statistics

Andy Williams — Mon, 13 Jun 2011 08:10:49 -0000

Hi,

What an excellent product!

Now that the vast majority of X86 servers are now NUMA it would be really useful if collectl could collect and report on NUMA statistics using (for example) information in /sys/devices/system/node/node*/meminfo. This would help in deciding whether to start using zone reclaim etc.

Cheers,

-Andy

Decode UID via NIS

lsmarden — Tue, 01 Mar 2011 01:06:07 -0000

We use NIS (Yellow Pages) at my site and unfortunately collectl only appears to look at the local password file (/etc/passwd) to decode UIDs into Usernames. Since the majority of the processes that I'm trying to monitor are user processes, and those are all started by individuals with NIS accounts, none of the Usernames are being decoded and I only see UIDs in the Process output.

If possible, it would be great to search the local password file first, and if the Username for a specific UID is not found, to then do a Yellow Pages search for the UID/Username.

Keep up the work. Great Product.

Porting collectl to other OSes

Anonymous — Sat, 09 Oct 2010 22:49:52 -0000

This is probably asking a lot, but I would like to know if it would be possible to get collectl running on windows XP. I have used the collectl utility on RedHat Linux. I would like to provide the same information for windows XP and I like the way collectl works.

WMI format

Anonymous — Sun, 31 Jan 2010 15:21:53 -0000

You're very close to WMI format. (Time series data). But WMI enable single row for all values at one time value (and I haven't seen it in Collectl. Would like to have a wild card input playback and one row per time value out. Filter would be like:
“\LogicalDisk(*)\Avg. Disk sec/Transfer”
“\LogicalDisk(*)\Disk Reads/sec”
“\Network Interface(*)\Packets Sent/sec”
“\Process(*)\% Processor Time”

add HP DL185 to envrules.std

Evan Felix — Wed, 30 Dec 2009 18:27:53 -0000

Please add these mapping rules to the evnrules.std for the HP DL185:

root -#dmidecode|grep -m1 'Product Name'
Product Name: ProLiant DL185 G5

[pre]
/CPU(\d) Diode/CPUTemp$1/
/Power Ambient/ATemp/
/Front Panel Temp/FPTemp/

/proc/meminfo:\'Mapped\' does not contain anon.mem. any more

Fred Ciesielski — Sat, 27 Oct 2007 09:08:28 -0000

Mark,

I believe there is no bug in SLES10 regarding the mapped memory; what has changed compared to older versions of the kernel (SLES9, 2.6.5 or RHEL4, 2.6.9), is a split between anonymous memory and file mapped memory, which were previously the 'mapped memory' (perhaps what is described at http://www.uwsg.iu.edu/hypermail/linux/kernel/0703.2/1688.html\). A new filed named 'AnonPages' is now available in /proc/meminfo.
What appears now as 'mapped memory' is in general of little interest for our purpose, but we definitely need to read the 'AnonPages' field, as it appears in the current /proc/meminfo file:

MemTotal: 8178968 kB
MemFree: 763544 kB
Buffers: 57264 kB
Cached: 7020516 kB
SwapCached: 0 kB
Active: 823256 kB
Inactive: 6356388 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 8178968 kB
LowFree: 763544 kB
SwapTotal: 2104504 kB
SwapFree: 2104380 kB
Dirty: 8812 kB
Writeback: 0 kB
AnonPages: 102060 kB
Mapped: 25720 kB
Slab: 191404 kB
CommitLimit: 6193988 kB
Committed_AS: 479844 kB
PageTables: 3884 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 4432 kB
VmallocChunk: 34359733935 kB
HugePages_Total: 0
HugePages_Free: 0
HugePages_Rsvd: 0
Hugepagesize: 2048 kB

I don't know what is best and have not tried yet to patch my collectl files yet, but as a first attempt I would just replace 'Mapped' everywhere with 'Anonpages' ! I don't know whether adding the file mapped contribution is really necessary, but after all, it is also defendable and would provide full backwards compatibility.

Fred.