fix: collect lock data when lslocks has null path values

[nagonzalez]

This PR supports collecting lslocks lock data even when path is null. These changes were validated on the following systems:

• CentOS 7.x
• Rocky 8.x

lslocks output from Rocky 8.x

lslocks -J -o +BLOCKER
{
   "locks": [
      {"command": "p4d_1_bin", "pid": "2013551", "type": "FLOCK", "size": null, "mode": "WRITE", "m": "0", "start": "0", "end": "0", "path": null, "blocker": "123"}
   ]
}

lslocks output on CentOS 7.x

COMMAND           PID  TYPE SIZE MODE  M      START        END PATH BLOCKER
p4d_1_bin       48134 FLOCK   0B WRITE 0          0          0        123

sample output in /p4/1/logs/monitor_metrics.log

2024-01-22 04:27:01 pid 10458, user , cmd , table None, blocked by pid 689, user svc-account1, cmd sync, args -q --parallel threads=4,min=1,minsize=1 //...@5554459
DEBUG 2024-01-22 04:27:01,891 monitor_metrics.py 318: Writing to metrics file: /depotdata/metrics/locks.prom
DEBUG 2024-01-22 04:27:01,891 monitor_metrics.py 319: Metrics: # HELP p4_locks_db_read Database read locks
# TYPE p4_locks_db_read gauge
p4_locks_db_read{serverid="server_id_value",sdpinst="1"} 0
# HELP p4_locks_db_write Database write locks
# TYPE p4_locks_db_write gauge
p4_locks_db_write{serverid="server_id_value",sdpinst="1"} 0
# HELP p4_locks_cliententity_read clientEntity read locks
# TYPE p4_locks_cliententity_read gauge
p4_locks_cliententity_read{serverid="server_id_value",sdpinst="1"} 0
# HELP p4_locks_cliententity_write clientEntity write locks
# TYPE p4_locks_cliententity_write gauge
p4_locks_cliententity_write{serverid="server_id_value",sdpinst="1"} 0
# HELP p4_locks_meta_read meta db read locks
# TYPE p4_locks_meta_read gauge
p4_locks_meta_read{serverid="server_id_value",sdpinst="1"} 0
# HELP p4_locks_meta_write meta db write locks
# TYPE p4_locks_meta_write gauge
p4_locks_meta_write{serverid="server_id_value",sdpinst="1"} 0
# HELP p4_locks_cmds_blocked cmds blocked by locks
# TYPE p4_locks_cmds_blocked gauge
p4_locks_cmds_blocked{serverid="server_id_value",sdpinst="1"} 1

[nagonzalez]

II believe this also may mitigate issue here: https://github.com/perforce/p4prometheus/issues/37

[nagonzalez]

Here's some sample telemetry captured:

grep 'blocked by pid' /p4/1/logs/monitor_metrics.log
2024-01-24 20:47:01 pid 134136, user service-account, cmd client, table None, blocked by pid 134022, user user1, cmd sync, args t:\depot\...#head
2024-01-24 20:47:01 pid 134135, user service-account, cmd client, table None, blocked by pid 134022, user user1, cmd sync, args t:\depot\...#head
2024-01-24 20:47:01 pid 134136, user service-account, cmd client, table None, blocked by pid 134022, user user1, cmd sync, args t:\depot\...#head
2024-01-24 20:47:01 pid 134135, user service-account, cmd client, table None, blocked by pid 134022, user user1, cmd sync, args t:\depot\...#head

[rcowham]

Hi - can you also update the test harness please? test_monitor_metrics.py

[nagonzalez]

yeah, definitely. Lemme' take a look. I'll reach out if I have any questions.

[nagonzalez]

@rcowham : test harness updated.

I guess better said, the harness already had good sample data but tests weren't passing. monitor_metrics.py was updated to make tests pass

[rcowham]

Can you check against latest - should have merged this in effectively, including the use of "sudo" where possible to improve output.