refactor the PDU management commands

[parmentelat]

goals

• to replace the hard-coded commands like on-jaguar and similar with something that is more scalable
• to take into account the second PDU-based devices in C007
• and so, to come up with a declarative way to define the configuration

expected behaviour

and so we will replace the rhubarbe-script thingy with a more adequate rhubarbe-pdu command; see expected behaviour below

returncode

generally speaking:

• 255 means there's a failure
• 0 means everything went well
• 1 is reserved for the status command and means the device is turned OFF

config files

a new file in /etc/rhubarbe/inventory-pdus.yaml will allow to describe the location and addresses of hardware

[parmentelat]

todo

• [x] replace the script main subcommand with pdu
• [x] write inventory_pdus.py
• [x] make code more asynchronous
• [x] remove most of the scripts under rhubarbe/scripts, except for pdu (this will maybe one day be written in async pexpect, but that's a separate issue)
• [x] write the logic for multi-inputs devices
• [x] setup the IPs for c007_top and c007_bottom
• [ ] test the logic for multi-inputs devices
• [ ] add optional verbosity; at the very least to show when a soft reset occurs with the OFF and RESET commands
• [x] write monitorpdus.py
• [x] consider rewriting the lower layer using async Python and SNMP rather that ssh that is very slow and does not seem to cope with multiple concurrent accesses
  NOTES:
  • [ ] the current code for pdu probe is still suboptimal as it does many back and forth's; it's still much better than before, but ...
  • [x] the SNMP variables to observe the outlet status (on or off) is read-only, so we had to resort to ssh for actually controlling the outlet state

if on the contrary the ssh option is preserved:

• [ ] check for tmpfiles leak in /tmp/pdus
• [ ] on the longer run maybe, rewrite the pdu shell in Python using pexpect that claims to have an async mode not relevant

[parmentelat]

revised examples: using version 5.2 or above, type

rhubarbe-pdu --help

[parmentelat]

as of 5.2.1 we have some of this available, i.e. the script subcommand has gone and one can obtain/set an individual status using rhubarbe-pdu status|on|off|reset the monitoring however still needs more work, although the logic seems fine, it could be that several ssh simultaneous connections just can't fly with these boxes so truly snmp looks like the right way to go as far as frequent monitoring is concerned

[parmentelat]

as of 5.2.6:

• the monitoring is done with SNMP only; can still be improved (particularly when probing a whole PDU) but it's much faster than the ssh version
• actually controlling the outlet states is still done using ssh; this too can probably be simplified to remove the use of expect

[parmentelat]

good enough for now, closing