genie_python: usage statistics

[DominicOram]

As a developer I would like to know who is using which parts of our API. This would be useful as:

• It will give us more confidence in knowing how changes to our API will affect users
• We can target demos of changes to the users that will see the difference the most
• We can get a good idea of what errors are occuring

Acceptance Criteria

• [ ] (Most of the time, see below) When a user calls a genie_python command the following is logged centrally:
  • [ ] The command called
  • [ ] The arguments it was called with
  • [ ] The instrument it was called from
  • [ ] The time
  • [ ] Any errors that occured
  • [ ] Time commands took to respond
• [ ] Developers are able to search this log in a nice way
• [ ] This is transparent to the user, e.g. if genie_python can't log this data it won't slow down or crash
• [ ] It is sufficient that we just catch most calls, e.g. we can ignore calls being made not on the NDX machine and even ignore those outside the GUI if easier

[John-Holt-Tessella]

Part one already sort of exists. All genie commands are logged to C:\Instrument\var\logs\genie_python which is readable from anywhere. However it isn't particularly easy to access or get statistics from.

[rerpha]

assigning myself as I think this would be a good work experience project.

[rerpha]

Some thoughts:

• Graylog should do all of the above and supports custom filters so we can add the instrument name and so on
• graypy is a nice library for this - can also be output into grafana if we need to see metrics on our grafana instance
• We should use the logging built-in for logging in genie python anyway, this would be a good opportunity to set it up and use it

[rerpha]

https://github.com/ISISComputingGroup/ibex_developers_manual/wiki/Graylog created

[rerpha]

https://github.com/ISISComputingGroup/IBEX/issues/6645 created following on from this

[rerpha]

Note that this functionality was not added to the simulation API as it wasn't logging beforehand. This is a replacement and addition to what was previously being logged.

[rerpha]

PRs:

• genie python: https://github.com/ISISComputingGroup/genie_python/pull/336
• IBEX: https://github.com/ISISComputingGroup/ibex_gui/pull/1330
• release notes: https://github.com/ISISComputingGroup/IBEX/pull/6646

To review:

• Spin up graylog instance using Docker
• Set up a GELF/UDP input on Graylog, found under System -> Inputs
• Use a command such as cset("text")
• Check it gets put into Graylog and \var\logs\genie_python log file

As a note, we can leave it in the code as it silently doesn't push if there is no graylog instance up, it should still log to file. in #6645 we should change this to point to a central instance.

[DominicOram]

• Can you make a breadcrumb for the wiki page please?
• Can you add the 2nd bullet above into wiki page? - When you do this can yo expand it to be a bit more of a step by step on how to get it started? E.g. mention the 127.0.01 issue and that when making the new input using just the default settings is fine. ~I couldn't get the graylog server working. I think it's started up correctly: But I can't get into the web interface: When you're back can we have a look at this and make sure the documentation is correct?~
• I also think that a system test would be good, if possible.
• Can you add a message to dev-changes about installing graypy when this merged?
• Can you add a ticket for generally improving what is logged, particularly:
  • Whether or not you're in simulation mode
  • Better error logging, e.g. if I do a cget on a block that doesn't exist it's not obvious from the graylog message that this is the case

[rerpha]

* Can you make a breadcrumb for the wiki page please?

* Can you add the 2nd bullet above into wiki page?
  I couldn't get the graylog server working. I think it's started up correctly:
  ![image](https://user-images.githubusercontent.com/10086797/126326068-43a567a3-9567-487e-91d7-693fe5a58a2f.png)
  But I can't get into the web interface:
  ![image](https://user-images.githubusercontent.com/10086797/126326155-3c196dd9-b6c6-4197-b5ea-00bd46d5dac1.png)
  When you're back can we have a look at this and make sure the documentation is correct?
  I also think that a system test would be good, if possible.

you need to use 127.0.0.1 instead of localhost, not sure if its a bug but i had the same issue. i think this could be a docker-port binding issue but it can be explicitly set when deployed.

will finish this ticket up when i have some time

[rerpha]

I tried to set up the inputs etc via a content pack but had no luck after a few hours of playing, so we'll have to manually do that on deployment - the wiki page should show instructions on how to add the inputs. I added a breadcrumb as well. will have a go at system testing now

[rerpha]

system testing is going to be a pain to do as we need to automatically set up inputs and extractors each time the graylog docker container is started which i haven't managed to successfully do

[DominicOram]

I added system testing to the creating the central graylog instance ticket