format for CPU data?
list Mike
I'm writing agents for other systems than linux and I want the data to be graphed just like the linux data is graphed. Can someone please post the expected format for CPU data so the data will be graphed? Mike
list Charles Goyard
Hi,
▸
Mike wrote:I'm writing agents for other systems than linux and I want the data to be graphed just like the linux data is graphed. Can someone please post the expected format for CPU data so the data will be graphed?
The real good way is to implement the native format of your OS on the server-side of hobbit. If it's a Unix-POSIX-like system, it's roughly a matter of cut and paste of the existing code. I did it for Sco's Unix (bleh!), and it made me understand hobbit internals much. Regards, -- Charles Goyard - user-98f9625a7a59@xymon.invalid - (+33) 1 45 38 01 31
list Thomas Kern
To get my z/VM system's CPU value to be reported by the 'STANDARD' bigbrother and hobbit servers, I send a message to the hobbit server with the text: status VMHOST,HQADMIN,DOE,GOV.cpu green ddd mmm dd hh:mm:ss tz yyyy up: nnnnn day(s), nnnn users, nnn procs,load=cpu% Then I can add a carriage return and any additional data I want. Then the hobbit page looks like this: Thu Oct 19 10:53:42 EDT 2006 up: 199 day(s), 1 users, 37 procs, load=2 z/VM Version 5 Release 1.0, service level 0402 (64-bit) Generated at 12/08/04 21:37:09 EDT IPL at 04/02/06 14:24:17 EDT ------------------------------------------------------------------------ -------- AVGPROC-002% 01 MDC READS-000001/SEC WRITES-000001/SEC HIT RATIO-097% STORAGE-053% PAGING-0001/SEC STEAL-000% Q0-00006(00000) DORMANT-00032 Q1-00000(00000) E1-00000(00000) Q2-00000(00000) EXPAN-001 E2-00000(00000) Q3-00000(00000) EXPAN-001 E3-00000(00000) PROC 0000-002% LIMITED-00000 /Thomas Kern /XXX-XXX-XXXX
▸
-----Original Message----- From: news [mailto:user-6269d477d278@xymon.invalid] On Behalf Of Mike Sent: Thursday, October 19, 2006 10:34 AM To: user-ae9b8668bcde@xymon.invalid Subject: [hobbit] format for CPU data? I'm writing agents for other systems than linux and I want the data to be graphed just like the linux data is graphed. Can someone please post the expected format for CPU data so the data will be graphed? Mike
list Thomas Kern
But then you get a column for VM-CPU, a column for MVS-CPU, a column for W2K-CPU, a column for WinXP-CPU, a column for Solaris-CPU (not really a linux), a column for AIX-CPU (another not really a linux) and finally a column for CPU (for the real linux systems). Now repeat that for DISK, MSGS, PROCS, FILES, MEMORY, PORTS, BACKUP, etc. You need a very WIDE screen when management wants it ALL on one page.
▸
/Thomas Kern
/XXX-XXX-XXXX -----Original Message----- From: Charles Goyard [mailto:user-98f9625a7a59@xymon.invalid] Sent: Thursday, October 19, 2006 10:55 AM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? Hi, Mike wrote:I'm writing agents for other systems than linux and I want the data to be graphed just like the linux data is graphed. Can someone please post the expected format for CPU data so the data will be graphed?The real good way is to implement the native format of your OS on the server-side of hobbit. If it's a Unix-POSIX-like system, it's roughly a matter of cut and paste of the existing code. I did it for Sco's Unix (bleh!), and it made me understand hobbit internals much. Regards, -- Charles Goyard - user-98f9625a7a59@xymon.invalid - (+33) 1 45 38 01 31
list Mike
▸
On 2006-10-19, Charles Goyard <user-98f9625a7a59@xymon.invalid> wrote:
Hi, Mike wrote:I'm writing agents for other systems than linux and I want the data to be graphed just like the linux data is graphed. Can someone please post the expected format for CPU data so the data will be graphed?The real good way is to implement the native format of your OS on the server-side of hobbit. If it's a Unix-POSIX-like system, it's roughly a matter of cut and paste of the existing code. I did it for Sco's Unix (bleh!), and it made me understand hobbit internals much. Regards,
The boxes I'm trying to get reporting are windows and I'm using vbscript. I'm happy to post the final client when it all works. Since I'm writing the client side there is not native format. I prefer the unix format as I'm a unix-head of many years. Mike
list Charles Goyard
Not at all. Here I have something like 5 or 6 Unix flavors, all are reporting under the same cpu/disk/procs/whatever. From the osname reported by the client, hobbitd can hand the data to the appropriate backend (it even has different backends for a few Linux flavor, such as RHEL3 or Debian). The idea it to put together a native backend for your OS.
▸
Kern, Thomas a écrit :But then you get a column for VM-CPU, a column for MVS-CPU, a column for W2K-CPU, a column for WinXP-CPU, a column for Solaris-CPU (not really a linux), a column for AIX-CPU (another not really a linux) and finally a column for CPU (for the real linux systems). Now repeat that for DISK, MSGS, PROCS, FILES, MEMORY, PORTS, BACKUP, etc. You need a very WIDE screen when management wants it ALL on one page.
-- Charles Goyard - user-98f9625a7a59@xymon.invalid - (+33) 1 45 38 01 31
list Rich Smrcina
There is a native backend for z/VM, it was integrated into Hobbit some time ago. It appears that Thomas wants to consolidate the CPU, Users and Procs into one column. I took a different approach, that since there was already a procs column reported by other clients, that the information that translated best to procs on z/VM was reported there. The concept of users and procs is somewhat blurred on z/VM, so I didn't differentiate the two. Some shops (it appears that his is one of them) have a clear demarkation of who a user is, based on their userid.
▸
Charles Goyard wrote:Not at all. Here I have something like 5 or 6 Unix flavors, all are reporting under the same cpu/disk/procs/whatever. From the osname reported by the client, hobbitd can hand the data to the appropriate backend (it even has different backends for a few Linux flavor, such as RHEL3 or Debian). The idea it to put together a native backend for your OS. Kern, Thomas a écrit :But then you get a column for VM-CPU, a column for MVS-CPU, a column for W2K-CPU, a column for WinXP-CPU, a column for Solaris-CPU (not really a linux), a column for AIX-CPU (another not really a linux) and finally a column for CPU (for the real linux systems). Now repeat that for DISK, MSGS, PROCS, FILES, MEMORY, PORTS, BACKUP, etc. You need a very WIDE screen when management wants it ALL on one page.
--
Rich Smrcina VM Assist, Inc. Phone: XXX-XXX-XXXX Ans Service: XXX-XXX-XXXX user-61add9955ef9@xymon.invalid Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
list Thomas Kern
Where in the CPU status message do I supply the z/VM 5.1 OSname & level information?
▸
/Thomas Kern
/XXX-XXX-XXXX -----Original Message----- From: Charles Goyard [mailto:user-98f9625a7a59@xymon.invalid] Sent: Thursday, October 19, 2006 11:22 AM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? Not at all. Here I have something like 5 or 6 Unix flavors, all are reporting under the same cpu/disk/procs/whatever. From the osname reported by the client, hobbitd can hand the data to the appropriate backend (it even has different backends for a few Linux flavor, such as RHEL3 or Debian). The idea it to put together a native backend for your OS. Kern, Thomas a écrit :But then you get a column for VM-CPU, a column for MVS-CPU, a column for W2K-CPU, a column for WinXP-CPU, a column for Solaris-CPU (not really a linux), a column for AIX-CPU (another not really a linux) and finally a column for CPU (for the real linux systems). Now repeat that for DISK, MSGS, PROCS, FILES, MEMORY, PORTS, BACKUP, etc. > > You need a very WIDE screen when management wants it ALL on one page.-- Charles Goyard - user-98f9625a7a59@xymon.invalid - (+33) 1 45 38 01 31
list Thomas Kern
Actually I do provide data to the PROCS test but I would like to provide it in a more 'STANDARD' format so that we don't have different columns for essentially the same data. What I would really like is to have the user count and proc count plotted properly, so I need to find the 'STANDARD' test parsing. /Thomas Kern /XXX-XXX-XXXX Thu Oct 19 11:43:43 EDT 2006 There are 38 users logged on. &green VMUTIL Ctime=3 09:28:53 Vtime=0 00:00:01 Ttime=0 00:00:01 IO=9295 &green PERFSVM Ctime=152 02:03:05 Vtime=0 02:03:50 Ttime=0 02:09:20 IO=1594094 &green TCPIP Ctime=12 16:51:29 Vtime=0 00:00:25 Ttime=0 00:00:36 IO=208 &green TCPHSF Ctime=199 21:03:23 Vtime=0 00:15:32 Ttime=0 00:29:51 IO=161 &green VSWCTRL1 Ctime=199 21:19:18 Vtime=0 00:00:11 Ttime=0 00:09:29 IO=188 &green VSWCTRL2 Ctime=199 21:19:18 Vtime=0 00:00:11 Ttime=0 00:13:56 IO=700
▸
-----Original Message----- From: Rich Smrcina [mailto:user-cf452ff334e0@xymon.invalid] Sent: Thursday, October 19, 2006 11:29 AM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? There is a native backend for z/VM, it was integrated into Hobbit some time ago. It appears that Thomas wants to consolidate the CPU, Users and Procs into one column. I took a different approach, that since there was already a procs column reported by other clients, that the information that translated best to procs on z/VM was reported there. The concept of users and procs is somewhat blurred on z/VM, so I didn't differentiate the two. Some shops (it appears that his is one of them) have a clear demarkation of who a user is, based on their userid. Charles Goyard wrote:Not at all.Here I have something like 5 or 6 Unix flavors, all are reporting under the same cpu/disk/procs/whatever.From the osname reported by the client, hobbitd can hand the data to the appropriate backend (it even has different backends for a few Linux flavor, such as RHEL3 or Debian). The idea it to put together a native backend for your OS.Kern, Thomas a écrit :But then you get a column for VM-CPU, a column for MVS-CPU, a column for W2K-CPU, a column for WinXP-CPU, a column for Solaris-CPU (not really a linux), a column for AIX-CPU (another not really a linux) and finally a column for CPU (for the real linux systems). Now repeat that for DISK, MSGS, PROCS, FILES, MEMORY, PORTS, BACKUP, etc. >> You need a very WIDE screen when management wants it ALL on one page.-- Rich Smrcina VM Assist, Inc. Phone: XXX-XXX-XXXX Ans Service: XXX-XXX-XXXX user-61add9955ef9@xymon.invalid Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
list Rich Smrcina
Then for procs you may need to present that data is 'ps -ef' format: UID PID PPID C STIME TTY TIME CMD root 1 0 0 Oct18 ? 00:00:01 init [5] root 2 1 0 Oct18 ? 00:00:00 [ksoftirqd/0] root 3 1 0 Oct18 ? 00:00:00 [events/0] root 4 1 0 Oct18 ? 00:00:00 [khelper] root 9 1 0 Oct18 ? 00:00:00 [kthread] root 19 9 0 Oct18 ? 00:00:08 [kacpid] root 116 9 0 Oct18 ? 00:00:00 [kblockd/0] root 156 9 0 Oct18 ? 00:00:00 [pdflush] Some of the data here doesn't really apply to z/VM, but it should be possible to come up with some sort of value for each column. UID and CMD would probably be the same thing, PID can be a counter, PPID would be 1, STIME would be the logon date or time, TTY is not relevant and TIME would be the CPUTime.
▸
Kern, Thomas wrote:Actually I do provide data to the PROCS test but I would like to provide it in a more 'STANDARD' format so that we don't have different columns for essentially the same data. What I would really like is to have the user count and proc count plotted properly, so I need to find the 'STANDARD' test parsing. /Thomas Kern /XXX-XXX-XXXX Thu Oct 19 11:43:43 EDT 2006 There are 38 users logged on. &green VMUTIL Ctime=3 09:28:53 Vtime=0 00:00:01 Ttime=0 00:00:01 IO=9295 &green PERFSVM Ctime=152 02:03:05 Vtime=0 02:03:50 Ttime=0 02:09:20 IO=1594094 &green TCPIP Ctime=12 16:51:29 Vtime=0 00:00:25 Ttime=0 00:00:36 IO=208 &green TCPHSF Ctime=199 21:03:23 Vtime=0 00:15:32 Ttime=0 00:29:51 IO=161 &green VSWCTRL1 Ctime=199 21:19:18 Vtime=0 00:00:11 Ttime=0 00:09:29 IO=188 &green VSWCTRL2 Ctime=199 21:19:18 Vtime=0 00:00:11 Ttime=0 00:13:56 IO=700-----Original Message----- From: Rich Smrcina [mailto:user-cf452ff334e0@xymon.invalid] Sent: Thursday, October 19, 2006 11:29 AM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? There is a native backend for z/VM, it was integrated into Hobbit some time ago. It appears that Thomas wants to consolidate the CPU, Users and Procs into one column. I took a different approach, that since there was already a procs column reported by other clients, that the information that translated best to procs on z/VM was reported there. The concept of users and procs is somewhat blurred on z/VM, so I didn't differentiate the two. Some shops (it appears that his is one of them) have a clear demarkation of who a user is, based on their userid. Charles Goyard wrote:Not at all. Here I have something like 5 or 6 Unix flavors, all are reporting under the same cpu/disk/procs/whatever.From the osname reported by the client, hobbitd can hand the data to the appropriate backend (it even has different backends for a few Linux flavor, such as RHEL3 or Debian). The idea it to put together a native backend for your OS.Kern, Thomas a écrit :But then you get a column for VM-CPU, a column forMVS-CPU, a column forW2K-CPU, a column for WinXP-CPU, a column for Solaris-CPU(not really alinux), a column for AIX-CPU (another not really a linux) and finally a column for CPU (for the real linux systems). Now repeat that for DISK, MSGS, PROCS, FILES, MEMORY, PORTS, BACKUP, etc. You need a very WIDE screen when management wants it ALL on one page.-- Rich Smrcina VM Assist, Inc. Phone: XXX-XXX-XXXX Ans Service: XXX-XXX-XXXX user-61add9955ef9@xymon.invalid Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
-- Rich Smrcina VM Assist, Inc. Phone: XXX-XXX-XXXX Ans Service: XXX-XXX-XXXX user-61add9955ef9@xymon.invalid Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
list Mike
▸
On 2006-10-19, Kern, Thomas <user-f1ebafb19faf@xymon.invalid> wrote:
To get my z/VM system's CPU value to be reported by the 'STANDARD' bigbrother and hobbit servers, I send a message to the hobbit server with the text: status VMHOST,HQADMIN,DOE,GOV.cpu green ddd mmm dd hh:mm:ss tz yyyy up: nnnnn day(s), nnnn users, nnn procs,load=cpu% Then I can add a carriage return and any additional data I want. Then the hobbit page looks like this: Thu Oct 19 10:53:42 EDT 2006 up: 199 day(s), 1 users, 37 procs, load=2 z/VM Version 5 Release 1.0, service level 0402 (64-bit) Generated at 12/08/04 21:37:09 EDT IPL at 04/02/06 14:24:17 EDT ------------------------------------------------------------------------ -------- AVGPROC-002% 01 MDC READS-000001/SEC WRITES-000001/SEC HIT RATIO-097% STORAGE-053% PAGING-0001/SEC STEAL-000% Q0-00006(00000) DORMANT-00032 Q1-00000(00000) E1-00000(00000) Q2-00000(00000) EXPAN-001 E2-00000(00000) Q3-00000(00000) EXPAN-001 E3-00000(00000) PROC 0000-002% LIMITED-00000 /Thomas Kern /XXX-XXX-XXXX-----Original Message----- From: news [mailto:user-6269d477d278@xymon.invalid] On Behalf Of Mike Sent: Thursday, October 19, 2006 10:34 AM To: user-ae9b8668bcde@xymon.invalid Subject: [hobbit] format for CPU data? I'm writing agents for other systems than linux and I want the data to be graphed just like the linux data is graphed. Can someone please post the expected format for CPU data so the data will be graphed? Mike
I get the status messages using the format 'status host-with-commas.test color message'. I want the CPU load averages to be graphed. Thank you for your format above, I cannot get hobbitd to graph the data using that format. Mike
list Thomas Kern
I just looked at the Hobbit server and it isn't plotting the cpu data either but the same format message is getting plotted by bigbrother. I need to break open that hobbit_rrd to find the parsing for these 'STANDARD' tests.
▸
/Thomas Kern
/XXX-XXX-XXXX
-----Original Message----- From: news [mailto:user-6269d477d278@xymon.invalid] On Behalf Of Mike Sent: Thursday, October 19, 2006 11:57 AM To: user-ae9b8668bcde@xymon.invalid Subject: [hobbit] Re: format for CPU data? I get the status messages using the format 'status host-with-commas.test color message'. I want the CPU load averages to be graphed. Thank you for your format above, I cannot get hobbitd to graph the data using that format. Mike
list Charles Goyard
▸
Kern, Thomas a écrit :
I just looked at the Hobbit server and it isn't plotting the cpu data either but the same format message is getting plotted by bigbrother. I need to break open that hobbit_rrd to find the parsing for these 'STANDARD' tests.
It can be that you have a rogue $HOBBITHOME/data/host/cpu.rrd with wrong DS. If so try to remove that file. -- Charles Goyard - user-98f9625a7a59@xymon.invalid - (+33) 1 45 38 01 31
list Thomas Kern
I will look at reworking the PROCS data to match 'ps -ef'. The user/procs count used to be obtained by bigbrother from the messages in the CPU test. CMD is irrelevant for most of the PROCS (service virtual machines) on z/VM, but for a real user, we could get the last command buffer. TTY could be filled in with the LUNAME, terminal device number or 'DISC' to be meaningful.
▸
/Thomas Kern
/XXX-XXX-XXXX -----Original Message----- From: Rich Smrcina [mailto:user-cf452ff334e0@xymon.invalid] Sent: Thursday, October 19, 2006 11:56 AM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? Then for procs you may need to present that data is 'ps -ef' format: UID PID PPID C STIME TTY TIME CMD root 1 0 0 Oct18 ? 00:00:01 init [5] root 2 1 0 Oct18 ? 00:00:00 [ksoftirqd/0] root 3 1 0 Oct18 ? 00:00:00 [events/0] root 4 1 0 Oct18 ? 00:00:00 [khelper] root 9 1 0 Oct18 ? 00:00:00 [kthread] root 19 9 0 Oct18 ? 00:00:08 [kacpid] root 116 9 0 Oct18 ? 00:00:00 [kblockd/0] root 156 9 0 Oct18 ? 00:00:00 [pdflush] Some of the data here doesn't really apply to z/VM, but it should be possible to come up with some sort of value for each column. UID and CMD would probably be the same thing, PID can be a counter, PPID would be 1, STIME would be the logon date or time, TTY is not relevant and TIME would be the CPUTime.
list Thomas Kern
I have no direct control over the hobbit server, but will forward this to that sysadmin.
▸
/Thomas Kern
/XXX-XXX-XXXX -----Original Message----- From: Charles Goyard [mailto:user-98f9625a7a59@xymon.invalid] Sent: Thursday, October 19, 2006 12:17 PM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] Re: format for CPU data? Kern, Thomas a écrit :I just looked at the Hobbit server and it isn't plotting the cpu data either but the same format message is getting plotted by bigbrother. I need to break open that hobbit_rrd to find the parsing for these 'STANDARD' tests.It can be that you have a rogue $HOBBITHOME/data/host/cpu.rrd with wrong DS. If so try to remove that file. -- Charles Goyard - user-98f9625a7a59@xymon.invalid - (+33) 1 45 38 01 31
list Rich Smrcina
The TTY idea makes alot of sense. Hopefully chasing down the last command isn't too much work (for the CPU that is), the one thing you don't want is for your network service monitoring to become a performance problem... :)
▸
Kern, Thomas wrote:I will look at reworking the PROCS data to match 'ps -ef'. The user/procs count used to be obtained by bigbrother from the messages in the CPU test. CMD is irrelevant for most of the PROCS (service virtual machines) on z/VM, but for a real user, we could get the last command buffer. TTY could be filled in with the LUNAME, terminal device number or 'DISC' to be meaningful. /Thomas Kern /XXX-XXX-XXXX-----Original Message----- From: Rich Smrcina [mailto:user-cf452ff334e0@xymon.invalid] Sent: Thursday, October 19, 2006 11:56 AM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? Then for procs you may need to present that data is 'ps -ef' format: UID PID PPID C STIME TTY TIME CMD root 1 0 0 Oct18 ? 00:00:01 init [5] root 2 1 0 Oct18 ? 00:00:00 [ksoftirqd/0] root 3 1 0 Oct18 ? 00:00:00 [events/0] root 4 1 0 Oct18 ? 00:00:00 [khelper] root 9 1 0 Oct18 ? 00:00:00 [kthread] root 19 9 0 Oct18 ? 00:00:08 [kacpid] root 116 9 0 Oct18 ? 00:00:00 [kblockd/0] root 156 9 0 Oct18 ? 00:00:00 [pdflush] Some of the data here doesn't really apply to z/VM, but it should be possible to come up with some sort of value for each column. UID and CMD would probably be the same thing, PID can be a counter, PPID would be 1, STIME would be the logon date or time, TTY is not relevant and TIME would be the CPUTime.
-- Rich Smrcina VM Assist, Inc. Phone: XXX-XXX-XXXX Ans Service: XXX-XXX-XXXX user-61add9955ef9@xymon.invalid Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
list Rich Smrcina
Ooohhh... lucky you! Kern, Thomas wrote:
For the number of REAL users on my z/VM systems, the cpu load of this monitoring should not be bad. And I can always try compiling the REXX code and EXECLOADing all of the pieces.
▸
/Thomas Kern /XXX-XXX-XXXX-----Original Message----- From: Rich Smrcina [mailto:user-cf452ff334e0@xymon.invalid] Sent: Thursday, October 19, 2006 12:21 PM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? The TTY idea makes alot of sense. Hopefully chasing down the last command isn't too much work (for the CPU that is), the one thing you don't want is for your network service monitoring to become a performance problem... :)
-- Rich Smrcina VM Assist, Inc. Phone: XXX-XXX-XXXX Ans Service: XXX-XXX-XXXX user-61add9955ef9@xymon.invalid Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
list Thomas Kern
For the number of REAL users on my z/VM systems, the cpu load of this monitoring should not be bad. And I can always try compiling the REXX code and EXECLOADing all of the pieces. /Thomas Kern /XXX-XXX-XXXX
-----Original Message----- From: Rich Smrcina [mailto:user-cf452ff334e0@xymon.invalid] Sent: Thursday, October 19, 2006 12:21 PM To: user-ae9b8668bcde@xymon.invalid Subject: Re: [hobbit] format for CPU data? The TTY idea makes alot of sense. Hopefully chasing down the last command isn't too much work (for the CPU that is), the one thing you don't want is for your network service monitoring to become a performance problem... :)