workload.googleapis.com/apache.current_connections not available on all platforms #1173
Closed as not planned
Description
I'm working on enabling more test coverage for apache, and have found that current_connections is not available on centos-7, sles-12, or sles-15. I believe it's because that metric is specific to the event Multi-Processing Module (MPM), and isn't present on the other MPMs, which are used on older distros. Here's a snapshot of the apache server status page for various distros. Note that some of them are missing ConnsTotal:
centos 7:
Total Accesses: 2
Total kBytes: 1
Uptime: 100
ReqPerSec: .02
BytesPerSec: 10.24
BytesPerReq: 512
BusyWorkers: 2
IdleWorkers: 4
Scoreboard: K_W___..........................................................................................................................................................................................................................................................
debian-11:
127.0.0.1
ServerVersion: Apache/2.4.56 (Debian)
ServerMPM: event
Server Built: 2023-03-08T03:05:04
CurrentTime: Wednesday, 22-Mar-2023 20:10:56 UTC
RestartTime: Wednesday, 22-Mar-2023 20:09:16 UTC
ParentServerConfigGeneration: 1
ParentServerMPMGeneration: 0
ServerUptimeSeconds: 100
ServerUptime: 1 minute 40 seconds
Load1: 0.21
Load5: 0.16
Load15: 0.07
Total Accesses: 2
Total kBytes: 1
Total Duration: 0
CPUUser: 0
CPUSystem: 0
CPUChildrenUser: 0
CPUChildrenSystem: 0
Uptime: 100
ReqPerSec: .02
BytesPerSec: 10.24
BytesPerReq: 512
DurationPerReq: 0
BusyWorkers: 1
IdleWorkers: 49
Processes: 2
Stopping: 0
BusyWorkers: 1
IdleWorkers: 49
ConnsTotal: 0
ConnsAsyncWriting: 0
ConnsAsyncKeepAlive: 0
ConnsAsyncClosing: 0
Scoreboard: ______________________________________W___________....................................................................................................
rocky-linux-8:
127.0.0.1
ServerVersion: Apache/2.4.37 (rocky)
ServerMPM: event
Server Built: Feb 21 2023 18:10:58
CurrentTime: Wednesday, 22-Mar-2023 20:17:13 UTC
RestartTime: Wednesday, 22-Mar-2023 20:15:20 UTC
ParentServerConfigGeneration: 1
ParentServerMPMGeneration: 0
ServerUptimeSeconds: 113
ServerUptime: 1 minute 53 seconds
Load1: 1.00
Load5: 1.19
Load15: 0.55
Total Accesses: 2
Total kBytes: 1
Total Duration: 0
CPUUser: .05
CPUSystem: .09
CPUChildrenUser: 0
CPUChildrenSystem: 0
CPULoad: .123894
Uptime: 113
ReqPerSec: .0176991
BytesPerSec: 9.06195
BytesPerReq: 512
DurationPerReq: 0
BusyWorkers: 1
IdleWorkers: 74
Processes: 3
Stopping: 0
BusyWorkers: 1
IdleWorkers: 74
ConnsTotal: 0
ConnsAsyncWriting: 0
ConnsAsyncKeepAlive: 0
ConnsAsyncClosing: 0
Scoreboard: __________________________W________________________________________________.....................................................................................................................................................................................................................................................................................................................................
sles-15:
127.0.0.1
ServerVersion: Apache/2.4.51 (Linux/SUSE) OpenSSL/1.1.1l
ServerMPM: prefork
Server Built: 2023-01-20 09:29:44.000000000 +0000
CurrentTime: Wednesday, 22-Mar-2023 20:18:24 UTC
RestartTime: Wednesday, 22-Mar-2023 20:16:45 UTC
ParentServerConfigGeneration: 1
ParentServerMPMGeneration: 0
ServerUptimeSeconds: 98
ServerUptime: 1 minute 38 seconds
Load1: 0.23
Load5: 0.28
Load15: 0.13
Total Accesses: 2
Total kBytes: 0
Total Duration: 3
CPUUser: .03
CPUSystem: .01
CPUChildrenUser: 0
CPUChildrenSystem: 0
CPULoad: .0408163
Uptime: 98
ReqPerSec: .0204082
BytesPerSec: 0
BytesPerReq: 0
DurationPerReq: 1.5
BusyWorkers: 2
IdleWorkers: 4
Scoreboard: K___W_................................................................................................................................................
My question is, should we disable testing for the current_connections metric on all distros or introduce a mechanism to disable specific metrics on specific distros?