| To: | pcp@xxxxxxxxxxx |
|---|---|
| Subject: | [Bug 1141] linux pmda does too much work for process queries |
| From: | bugzilla-daemon@xxxxxxxxxxx |
| Date: | Fri, 08 Apr 2016 14:11:17 +0000 |
| Auto-submitted: | auto-generated |
| Delivered-to: | pcp@xxxxxxxxxxx |
| In-reply-to: | <bug-1141-835@xxxxxxxxxxxxxxxx/bugzilla/> |
| References: | <bug-1141-835@xxxxxxxxxxxxxxxx/bugzilla/> |
|
Comment # 1
on bug 1141
from Frank Ch. Eigler
For example, the three proc.* metrics commonly included in default pmlogger
configurations (proc.runq.blocked, proc.runq.runnable, proc.nprocs) require
little to no actual /proc/*/{stat,cmdline} traffic. The values in /proc/stats
and /proc/loadavg, and possibly a directory enumeration of /proc/* (to count
processes vs. threads) should satisfy the metrics. Instead the pmda goes
through a "refresh", checking per-process stats and command lines and stuff.
Unnecessary.
You are receiving this mail because:
|
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | [Bug 1323521] remote operation of pmie based pmda restarter interferes with local pmcd, bugzilla |
|---|---|
| Next by Date: | [Bug 1325363] New: multithreaded clients with concurrent pmNewContext suffer deadlocks or low performance, bugzilla |
| Previous by Thread: | [Bug 1141] New: linux pmda does too much work for process queries, bugzilla-daemon |
| Next by Thread: | [Bug 1067] linux pmda does too much work for network.interface queries, bugzilla-daemon |
| Indexes: | [Date] [Thread] [Top] [All Lists] |