[checkmk-commits] 6735 FIX cpu_util_unix: Fixed calculation of CPU usage on UNIX

Module: check_mk Branch: master Commit: 79e551186b02cfbdbc1a9f681a6c07ac231984a8 URL: http://git.mathias-kettner.de/git/?p=check_mk.git;a=commit;h=79e551186b02cf… Author: Simon Betz &lt;si(a)mathias-kettner.de&gt; Date: Thu Sep 27 16:47:59 2018 +0200 6735 FIX cpu_util_unix: Fixed calculation of CPU usage on UNIX Thanks to Juergen Raschke for his patch! The result of the cpu_usage is wrong. The cpu_time_guest is included in cpu_time_user: See https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/kerne… for more information: <code> /* * Account guest CPU time to a process. * @p: the process that the CPU time gets accounted to * @cputime: the CPU time spent in virtual machine since the last update */ void account_guest_time(struct task_struct *p, u64 cputime) { u64 *cpustat = kcpustat_this_cpu->cpustat; /* Add guest time to process. */ p->utime += cputime; account_group_user_time(p, cputime); p->gtime += cputime; /* Add guest time to cpustat. */ if (task_nice(p) > 0) { cpustat[CPUTIME_NICE] += cputime; cpustat[CPUTIME_GUEST_NICE] += cputime; } else { cpustat[CPUTIME_USER] += cputime; cpustat[CPUTIME_GUEST] += cputime; } } </code> FEED-935 Change-Id: Ic26ae1d06da1284b61649831e4967776f0a89684 --- .werks/6735 | 40 ++++++++++++++++++++++++++++++++++++++++ checks/cpu_util.include | 16 +++++++++++++--- 2 files changed, 53 insertions(+), 3 deletions(-) diff --git a/.werks/6735 b/.werks/6735 new file mode 100644 index 0000000..4b5ed95 --- /dev/null +++ b/.werks/6735 @@ -0,0 +1,40 @@ +Title: cpu_util_unix: Fixed calculation of CPU usage on UNIX +Level: 1 +Component: checks +Class: fix +Compatible: compat +Edition: cre +State: unknown +Version: 1.6.0i1 +Date: 1538059527 + +The result of the cpu_usage is wrong. The cpu_time_guest is included in +cpu_time_user: See +https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/kernel/sched/cputime.c +for more information: + +<code> +/* + * Account guest CPU time to a process. + * @p: the process that the CPU time gets accounted to + * @cputime: the CPU time spent in virtual machine since the last update + */ +void account_guest_time(struct task_struct *p, u64 cputime) +{ + u64 *cpustat = kcpustat_this_cpu->cpustat; + + /* Add guest time to process. */ + p->utime += cputime; + account_group_user_time(p, cputime); + p->gtime += cputime; + + /* Add guest time to cpustat. */ + if (task_nice(p) > 0) { + cpustat[CPUTIME_NICE] += cputime; + cpustat[CPUTIME_GUEST_NICE] += cputime; + } else { + cpustat[CPUTIME_USER] += cputime; + cpustat[CPUTIME_GUEST] += cputime; + } +} +</code> diff --git a/checks/cpu_util.include b/checks/cpu_util.include index c5557cc..eb429af 100644 --- a/checks/cpu_util.include +++ b/checks/cpu_util.include @@ -132,8 +132,8 @@ def check_cpu_util(util, params, this_time=None, cores=None): # - 5 - irq: servicing interrupts # - 6 - softirq: servicing softirqs # - 7 - steal: involuntary wait -# - 8 - guest: time spent in guest OK -# - 9 - guest_nice: time spent in niced guest OK +# - 8 - guest: time spent in guest OK, also counted in 0 (user) +# - 9 - guest_nice: time spent in niced guest OK, also counted in 1 (nice) def check_cpu_util_unix(values, params, cores = None): this_time = time.time() # Compute jiffi-differences of all relevant counters @@ -150,7 +150,17 @@ def check_cpu_util_unix(values, params, cores = None): if sum_jiffies == 0: raise MKCounterWrapped("Too short time difference since last check") - user = diff_values[0] + diff_values[1] # add user + nice + user = diff_values[0] + diff_values[1] # add user + nice + if len(values) >= 10: # and (values[8] or values[9]): + # https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/kerne… + # see 'account_guest_time' + # if task_nice(p) <= 0: + # cpustat[CPUTIME_USER] += cputime; + # cpustat[CPUTIME_GUEST] += cputime; + subtrahend = diff_values[8] + diff_values[9] + user -= subtrahend + sum_jiffies -= subtrahend + system = diff_values[2] + diff_values[5] + diff_values[6] wait = diff_values[4] user_perc = 100.0 * float(user) / float(sum_jiffies)