gopsutil
gopsutil copied to clipboard
shirou
[aix][host]: Uptime might have a comma
#1809 accidentally introduced some of bug on AIX. (see this comment)
This PR fixes it and add , to all of other symbols.
Note: I don't have a AIX machine, so I can not check this works or not.
@Dylan-M sorry to late fix. But could you check this PR is actually you want? Thanks!
@Dylan-M Could you provide these uptime examples like in the test? I want to add those to the test.
- additional commas
- ut[5] and/or ut[7] is missing
@Dylan-M Could you provide these uptime examples like in the test? I want to add those to the test.
* additional commas * ut[5] and/or ut[7] is missing
I can't provide them easily, because it requires getting the system that shows those formats - since they only show when they're non-zero.
Also, to further complicate it, one of my systems outputs uptime like this:
# uptime
10:58AM up 22:28, 2 users, load average: 1.87, 1.96, 1.96
Which is for less than a full day, but doesn't split it hrs and mins like I expected. I'll check back once it has been running > 24 hours to see what that looks like. I really hate the way AIX does uptime :joy:
@Dylan-M Could you provide these uptime examples like in the test? I want to add those to the test.
* additional commas * ut[5] and/or ut[7] is missingI can't provide them easily, because it requires getting the system that shows those formats - since they only show when they're non-zero.
Also, to further complicate it, one of my systems outputs uptime like this:
# uptime 10:58AM up 22:28, 2 users, load average: 1.87, 1.96, 1.96Which is for less than a full day, but doesn't split it hrs and mins like I expected. I'll check back once it has been running > 24 hours to see what that looks like. I really hate the way AIX does uptime 😂
Here is what I have recorded (well, partial, I'm recording to a file every 15 minutes using a cronjob)
11:45AM up 23:14, 2 users, load average: 2.79, 2.72, 2.53
12:00PM up 23:29, 2 users, load average: 3.01, 2.87, 2.75
12:15PM up 23:44, 2 users, load average: 2.02, 1.99, 2.26
12:30PM up 23:59, 2 users, load average: 2.73, 2.50, 2.40
12:45PM up 1 day, 14 mins, 2 users, load average: 2.15, 2.31, 2.40
01:00PM up 1 day, 29 mins, 2 users, load average: 2.66, 2.60, 2.44
01:15PM up 1 day, 44 mins, 2 users, load average: 2.01, 2.31, 2.45
01:30PM up 1 day, 59 mins, 2 users, load average: 2.00, 2.09, 2.26
01:45PM up 1 day, 1:14, 2 users, load average: 2.14, 1.97, 2.06
02:00PM up 1 day, 1:29, 2 users, load average: 3.04, 2.56, 2.25
02:15PM up 1 day, 1:44, 2 users, load average: 2.94, 2.89, 2.67
02:30PM up 1 day, 1:59, 2 users, load average: 2.08, 2.07, 2.29
And from later, another day crossing:
11:45AM up 1 day, 23:14, 2 users, load average: 1.84, 2.12, 2.30
12:00PM up 1 day, 23:29, 2 users, load average: 2.63, 2.19, 2.18
12:15PM up 1 day, 23:44, 2 users, load average: 3.13, 2.58, 2.30
12:30PM up 1 day, 23:59, 2 users, load average: 2.94, 2.48, 2.29
12:45PM up 2 days, 14 mins, 2 users, load average: 2.83, 2.75, 2.58
01:00PM up 2 days, 29 mins, 2 users, load average: 2.80, 2.76, 2.64
01:15PM up 2 days, 44 mins, 2 users, load average: 1.75, 2.07, 2.38
01:30PM up 2 days, 59 mins, 2 users, load average: 2.93, 2.36, 2.24
01:45PM up 2 days, 1:14, 2 users, load average: 2.21, 2.27, 2.37
02:00PM up 2 days, 1:29, 2 users, load average: 1.78, 2.38, 2.49
02:15PM up 2 days, 1:44, 2 users, load average: 2.87, 2.44, 2.38
02:30PM up 2 days, 1:59, 2 users, load average: 2.80, 2.46, 2.33
@Dylan-M Thank you for the additional information. This is more complex than I expected. I’m starting to feel that a proper fix might be better than a quick workaround. It might be better to make use of regular expressions. I'm a busy for these days, so it might take me some time, but I’ll give it a try. Alternatively, feel free to send a good update yourself if you’d like!
@Dylan-M Thank you for the additional information. This is more complex than I expected. I’m starting to feel that a proper fix might be better than a quick workaround. It might be better to make use of regular expressions. I'm a busy for these days, so it might take me some time, but I’ll give it a try. Alternatively, feel free to send a good update yourself if you’d like!
I'm sure one of us will get to it when we have time. I'm swamped with my day job work right now too.
I think I've found a much better approach than the uptime command (and it's poorly machine readable format):
Get the start time of the init process (pid 1)
ps -o etime -p 1
Here is the man page info for the etime formatting option
etime
Indicates the elapsed time since the process started. The elapsed time is displayed in the following format:
[[ dd-]hh:]mm:ss
This should be significantly easier to parse.
My above solution using ps, also seems significantly more accurate (uptime doesn't properly roll the minute accurately)
[root@aix72 ~]# uptime
11:37AM up 15:02, 2 users, load average: 2.97, 2.94, 2.76
[root@aix72 ~]# ps -o etimes -p 1
ELAPSED
15:03:02
[root@aix72 ~]# ps -o etimes -p 1
ELAPSED
15:03:13
[root@aix72 ~]# uptime
11:37AM up 15:02, 2 users, load average: 3.12, 2.97, 2.78
[root@aix72 ~]# ps -o etimes -p 1
ELAPSED
15:03:17
We can also pipe to fgrep to remove the ELPASED header more efficiently than we can probably do in Golang code.
[root@aix72 ~]# ps -o etimes -p 1 | fgrep -v ELAPSED
15:04:34
Looking at this: https://www.ibm.com/docs/en/aix/7.3.0?topic=p-ps-command
It does seem that etime (or etimes? which is correct?) uses a fixed format [[ dd-]hh:]mm:ss, looks good! Since using a pipe would involve spawning two processes, handling this level of string processing directly in Go is probably faster and easier.
(On Linux, there's a --no-headers flag, but it seems the AIX ps command doesn't have that.)
Circling back to this (AIX work in this library in general) next week. I'll handle fixing this as part of that.
As for etime vs etimes - apparently both work and are valid. Yay oddities. I'll double check for support across multiple versions when I implement it.