cpu usage reported by port 311

Question

cpu usage reported by port 311

I am using Nagios to monitor my servers (one Linux box and one Mac Mini Server with Mountain Lion, all upgrades installed).

For the Mac server, I was using check_osx_server plugin, which is using port 311 to get information from the OS-X server. After the last upgrade (I am not sure if it started with the 10.8.2 upgrade or the latest supplemental upgrade), the CPU Usage reported by the server is very strange. The thing is, it's not Nagios which is reporting wrong numbers, it is the server itself which is returning erroneous values. You can check this by opening a browser window and point it to https://yourserver.com:311 and then select servermgr_info and then from the drop-down menu getHardwareInfo and click Send Command. That XML document returned from the server, is where check_osx_server plugin is getting its numbers from.

What I have found so far:

1. When you first power up the server, if you check the CPU usage you get a number with 20-something digits. This is crazy, because the number reported is supposed to be less than 10000, since that number is divided by 100, to show you the CPU Usage percentage.

2. The number remains the same for ever. Obviously, the CPU Usage changes with time, and that was reported accordingly, in the past. Now this number changes only if you run Server.app in the server or on a remote Mac. It changes to a more meaningful number (four digits, since this number has to be divided by 100, to get the CPU Usage percentage), but again, it remains the same for ever, until you close Server.app and re-run it. It also changes, if you go to the Stats in Server.app and change the default period from 4 hours, to 1 hour, but not if you change from one hour to 4 or 24h. Here is a sample of the XML file returned:

> <?xml version="1.0" encoding="UTF-8"?>

> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN"

> "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0">

> <dict>

> <key>command</key>

> <string>getHardwareInfo</string>

> <key>cpuUsageBy100</key>

> <integer>9183</integer>

> <key>cpuUsagesBy100Array</key>

> <array>

> <dict>

> <key>cpuUsage-0</key>

> <integer>9094</integer>

> </dict>

> <dict>

> <key>cpuUsage-1</key>

> <integer>9272</integer>

> </dict>

> </array>

There are more lines, but I have cut the file for brevity.

3. Again, as long as the Server.app is running, the CPU Usage reported does NOT change. Only if you quit Server.app and re-run it, it changes.

From the above, it is clear that port 311 is no longer reporting the momentary CPU Usage, but some other value, which is stored somewhere and is not updated unless you run Server.app.

Has anyone else noticed this behavior?

Posted on Oct 11, 2012 4:25 AM

Reply

Answer 1

Oct 21, 2012 5:51 AM in response to John Caradimas

Can someone who has access to a Mac Mini Server running Mountain Lion, try this? Just point a browser to your server port 311 (https://yourserver.com:311) and select servermgr_info and then from the drop-down menu getHardwareInfo and click Send Command. Then notice the number below <key>cpuUsageBy100</key> and then refresh the page (do a reload). Does the number change for you??

Reply

Answer 2

Nov 21, 2012 1:51 PM in response to John Caradimas

I had this occansionally on 10.7.5 and the only way to get rid of the stucked CPU number was to go to https://myserver:311 and sending the corresponding command. this seamed to refresh everything.

Reply

Answer 3

Nov 21, 2012 2:44 PM in response to henningkessler

Yes, that's what was happening with 10.7.xx. That doesn't happen with 10.8.xx though.

Reply

Answer 4

Nov 21, 2012 2:47 PM in response to John Caradimas

Okay tomorrow i probably will set up the libsrvrmgrd-osx on a 10.8.2 server. sounds promising :-)

Reply

Answer 5

Nov 21, 2012 2:51 PM in response to henningkessler

I would appreciate if you could let me know of your findings.

Reply

Answer 6

Nov 23, 2012 1:03 PM in response to John Caradimas

tried but stuck funnily with this message on every check:

ERROR:ProblemContactingServer:<urlopen error [Errno8] _ssl.c:504: EOF occurred in violation of protocol>

until i fixed this issue i probably can't say anything...

Reply

Answer 7

Nov 23, 2012 1:17 PM in response to henningkessler

Why don't you just try using a web browser window, pointed to https://yourserver.com:311?

Reply

Answer 8

Nov 23, 2012 1:55 PM in response to John Caradimas

Okay, i can confirm your experiences. the only thing i have not done ist to restart the server.

Seams to be bug, but probably apple won't fix this as with the end of severadmin in 10.8 this method of pulling informations could be seen as depcrecated :-(

Reply

Answer 9

Nov 23, 2012 3:42 PM in response to henningkessler

I am afraid you are right. Unfortunately, the new "Server.app" is nowhere near being an industry standard, so I wonder how will sysadmins where Mac Servers coexist with other machines will manage their servers farm. I mean, with that serveradmin interface, you could monitor your Mac servers in the same way you monitor your Linux machines. Now, how can one monitor a Mac in a mixed environment is anyone's guess. Something tells me that Apple is no longer interested in being a server manufacturer, which means that sooner or later my poor Mac Mini Server will be delegated to other tasks.

Thanks for confirming my findings, I was very surprised that no one has noticed this so far.

Reply

Answer 10

Nov 24, 2012 1:17 PM in response to John Caradimas

Well lets see how things are developing...

But if your issues are about how to monitor your mac servers than you should probably have look at SNMP as very capable possibility, as you can integrate it very well into nagios and even extend it with own scripts running some nice severadmin commands and grap information from that.

http://www.bynkii.com/archives/2007/12/a_primer_to_snmp_on_mac_os_x_1.html

Reply

Answer 11

Nov 24, 2012 2:05 PM in response to henningkessler

I'll check your link later tomorrow, but I am not sure if SNMP can monitor the same services as libsrvrmgrd-osx can via Nagios and the port 311. Right now, I have a one-page monitoring report, which shows me everything I need to know about all my routers and servers. I have no idea how I can replicate that via SNMP.

I'll get back after I check your link.

Again, many thanks!

Reply

Answer 12

Apr 18, 2013 7:25 AM in response to John Caradimas

I know this is an older thread but I thought that I would reply with my findings.

John is correct, you need to open Server.app to unstick the cpu integer, then it will change once and stick again until you close and re-open the app again.

I haven't found a way to get it to refresh properly yet, SNMP might be the best way around it.

Interestingly enough on a 10.8.3 server I have just installed I get the error below (exactly the same as above). Same command returns correctly on a server running 10.7.4 so I guess its a 10.8.3 thing.

ERROR: Problem Contacting Server:<urlopen error [Errno 8] _ssl.c:490: EOF occurred in violation of protocol>

Reply

Answer 13

Apr 18, 2013 7:33 AM in response to Billygunn

sounds like a bug i had before. Probably with some newer prerelease versions of the plugin your problem will be solved. check out the discussion on http://code.google.com/p/libsrvrmgrd-osx/issues/detail?id=16.

Reply

Answer 14

Apr 18, 2013 8:30 AM in response to henningkessler

Thanks for the tip, but replacing the files hasn't helped.

I'll post on their thread.

Reply

Answer 15

Apr 18, 2013 8:35 AM in response to Billygunn

I do not think you will get any solution. The issue is not the various add-ons, the server is NOT updating the information any more. It's an Apple issue.

Reply