Missing graphs with ES 0.20.2 about bigdesk

[lukas-vlcek]

ok, I will have a look. Did you only upgrad the ES (from which version?) or did you also make more changes? For example did you migrated to new HW, switched to different user account that is running ES, switched hosting provider... etc?

One option is that the ES JSON response format has changed. The other option is that you are missing sigar libs or the user does not have permissions to read the system info on that machine.

[electrical]

Hiya,

I've only upgraded ES to 0.20.2 from 0.19.x ( with a complete restart of course )

Strange thing is that on a fresh 0.20.2 machine its working fine.

[lukas-vlcek]

Does it mean you are able to replicate the issue?

[Alex-Ikanow]

Did you ever work out what happened here?

I have the same problem (ie exact same graphs greyed out, same ones working) on 0.18 and 0.19 (I recently upgraded from an old bigdesk version for 0.18 on which it wasn't very noticeable)

Is this a permissions issues? I tried reading some values of "/proc" as (non-root) user elasticsearch, eg:

runuser -s /bin/bash - elasticsearch -c "cat /proc/meminfo"

and it worked fine, but I'm not sure how exactly you are collecting the system stats?

(I'm running on a mix of Centos5/Centos6 machines on Amazon EC2 if it makes any difference)

Thanks for any insight/debugging suggestions!

[lukas-vlcek]

This system info is collected via sigar library. And I can think of basically the following reasons why it may not work:

• sigar library is not provided or is missing on the classpath
  • For example I think sigar is not provided by default when ES is started in embedded mode
• sigar does not support your platform
  • Consult http://support.hyperic.com/display/SIGAR/Home for compatibility matrix
  • Also check that libraries that are required to support your platform are found in <ES_HOME>/lib/sigar
• sigar libs fail to load
  • May be you can try to turn on low level logging to learn about this (see [1,2])
• sigar is loaded but elasticsearch user does not have a permissions to read this system level info or the read fails due to some other reason
  • Not sure if this can really happen (may be it would fail to initialize the libs in first place) but according to the code in such case the related SigarExceptions are not logged [3] (not saying this is ES bug!)
  • Try to use curl to get the raw JSON data from the admin API to see if there are any data available that should be displayed in the charts in the bigdesk

[1] As of writing the latest 0.19.12 ES load sigar libs in the following code:
https://github.com/elasticsearch/elasticsearch/blob/v0.19.12/src/main/java/org/elasticsearch/monitor/sigar/SigarService.java
So you can turn on appropriate log at the trace level to see more details.

[2] Sigar fail can also be logged in MonitorModule configuration, see:
https://github.com/elasticsearch/elasticsearch/blob/v0.19.12/src/main/java/org/elasticsearch/monitor/MonitorModule.java#L80 (interestingly, it can fall back to JMX system info collector implementation)

[3] For example check Network info: https://github.com/elasticsearch/elasticsearch/blob/v0.19.12/src/main/java/org/elasticsearch/monitor/network/SigarNetworkProbe.java

HTH, basically I did not personally find me in the situation where the charts were missing so I did not have an urgent need to think about possible workarounds. But I wouldn't mind improving bigdesk if I learn how in this case. So feedback is really welcome.

[Alex-Ikanow]

OK thanks for the pointers - one(-ish) more dumb question ... can you give me an example of how you're getting the data for one of the charts I can't display?

Eg which elasticsearch call if that's how you're getting all of them? There's a load of status calls in the API, they all seem to work but maybe they should be returning less than they are (and hence the resulting graphs are blank)

eg under OS, I not only get no graphs, but also the following text:
CPU vendor: n/a
CPU model: n/a
CPU total logical cores: n/a
CPU cache: n/a
which API call does that come from?

Similarly under Filesystem I get no graphs (read/writes are "n/a"), and device and mount are "n/a" (the path/free/available/total are correctly populated) - which API call do these come from?

Thanks again!

Alex

[Alex-Ikanow]

Heh never mind, looks like my install RPM doesn't copy the .so files across, I suppose I'm truly a java programmer now when my first thought isn't "gah must be some problem with shared object files", not sure if that's a good thing or not.. :)

I guess you can close this issue since the OP must have had the same problem based on his description?

Rudely late edit: thanks again for the pointers!

Even more rudely late edit: also thanks for providing an a great utility! (even more awesome now I have all these "extra" graphs)

[electrical]

I think we can close this issue.

[lukas-vlcek]

Thanks @electrical