diff --git a/profiles/openshift/tuned.conf b/profiles/openshift/tuned.conf index 197850c5..6c30c2a4 100644 --- a/profiles/openshift/tuned.conf +++ b/profiles/openshift/tuned.conf @@ -33,4 +33,4 @@ vm.max_map_count=262144 # see rhbz#1979352; exclude containers from aligning to house keeping CPUs cgroup_ps_blacklist=/kubepods\.slice/ # workaround for rhbz#1921738 -runtime=0 +dynamic=0 diff --git a/tests/beakerlib/Traceback-caused-by-scheduler-plugin-with/runtest.sh b/tests/beakerlib/Traceback-caused-by-scheduler-plugin-with/runtest.sh index d5cce712..5a13b7f0 100755 --- a/tests/beakerlib/Traceback-caused-by-scheduler-plugin-with/runtest.sh +++ b/tests/beakerlib/Traceback-caused-by-scheduler-plugin-with/runtest.sh @@ -43,7 +43,7 @@ rlJournalStart rlRun "pushd $PROFILE_DIR/test-profile" cat << EOF > tuned.conf [scheduler] -runtime=0 +dynamic=0 EOF rlRun "popd" diff --git a/tuned-main.conf b/tuned-main.conf index 1f1269f8..edbd49e4 100644 --- a/tuned-main.conf +++ b/tuned-main.conf @@ -7,7 +7,12 @@ daemon = 1 # Dynamically tune devices, if disabled only static tuning will be used. -dynamic_tuning = 0 +dynamic_tuning = 1 + +# A list of plugins to allow dynamic tuning for. The "dynamic_tuning" +# option must be enabled for this to have any effect. If not set, all +# plugins are allowed to perform dynamic tuning. +dynamic_plugins = scheduler # How long to sleep before checking for events (in seconds) # higher number means lower overhead but longer response time. diff --git a/tuned/plugins/plugin_scheduler.py b/tuned/plugins/plugin_scheduler.py index 171faa98..64cf0cbe 100644 --- a/tuned/plugins/plugin_scheduler.py +++ b/tuned/plugins/plugin_scheduler.py @@ -273,7 +273,7 @@ class SchedulerPlugin(base.Plugin): processes are not processed by the scheduler plug-in. + The CPU overhead of the scheduler plugin can be mitigated by using - the scheduler [option]`runtime` option and setting it to `0`. This + the scheduler [option]`dynamic` option and setting it to `0`. This will completely disable the dynamic scheduler functionality and the perf events will not be monitored and acted upon. The disadvantage ot this approach is the procees/thread tuning will be done only at @@ -283,7 +283,7 @@ class SchedulerPlugin(base.Plugin): ==== ---- [scheduler] - runtime=0 + dynamic=0 isolated_cores=1,3 ---- ==== @@ -465,6 +465,10 @@ def supports_static_tuning(cls): @classmethod def supports_dynamic_tuning(cls): + return True + + @classmethod + def uses_periodic_tuning(cls): return False def _calc_mmap_pages(self, mmap_pages): @@ -482,10 +486,6 @@ def _calc_mmap_pages(self, mmap_pages): def _instance_init(self, instance): super(SchedulerPlugin, self)._instance_init(instance) instance._evlist = None - # this is hack, runtime_tuning should be covered by dynamic_tuning configuration - # TODO: add per plugin dynamic tuning configuration and use dynamic_tuning configuration - # instead of runtime_tuning - instance._runtime_tuning = True # FIXME: do we want to do this here? # recover original values in case of crash @@ -510,42 +510,44 @@ def _instance_init(self, instance): instance._scheduler = instance.options perf_mmap_pages_raw = self._variables.expand(instance.options["perf_mmap_pages"]) - perf_mmap_pages = self._calc_mmap_pages(perf_mmap_pages_raw) - if perf_mmap_pages == 0: + instance._perf_mmap_pages = self._calc_mmap_pages(perf_mmap_pages_raw) + if instance._perf_mmap_pages == 0: log.error("Invalid 'perf_mmap_pages' value specified: '%s', using default kernel value" % perf_mmap_pages_raw) - perf_mmap_pages = None - if perf_mmap_pages is not None and str(perf_mmap_pages) != perf_mmap_pages_raw: + instance._perf_mmap_pages = None + if instance._perf_mmap_pages is not None and str(instance._perf_mmap_pages) != perf_mmap_pages_raw: log.info("'perf_mmap_pages' value has to be power of two, specified: '%s', using: '%d'" % - (perf_mmap_pages_raw, perf_mmap_pages)) + (perf_mmap_pages_raw, instance._perf_mmap_pages)) for k in instance._scheduler: instance._scheduler[k] = self._variables.expand(instance._scheduler[k]) - if self._cmd.get_bool(instance._scheduler.get("runtime", 1)) == "0": - instance._runtime_tuning = False - instance._terminate = threading.Event() - if self._daemon and instance._runtime_tuning: - try: - instance._threads = perf.thread_map() - evsel = perf.evsel(type = perf.TYPE_SOFTWARE, - config = perf.COUNT_SW_DUMMY, - task = 1, comm = 1, mmap = 0, freq = 0, - wakeup_events = 1, watermark = 1, - sample_type = perf.SAMPLE_TID | perf.SAMPLE_CPU) - evsel.open(cpus = self._cpus, threads = instance._threads) - instance._evlist = perf.evlist(self._cpus, instance._threads) - instance._evlist.add(evsel) - if perf_mmap_pages is None: - instance._evlist.mmap() - else: - instance._evlist.mmap(pages = perf_mmap_pages) - # no perf - except: - instance._runtime_tuning = False def _instance_cleanup(self, instance): if instance._evlist: for fd in instance._evlist.get_pollfd(): os.close(fd.name) + def _instance_init_dynamic(self, instance): + super(SchedulerPlugin, self)._instance_init_dynamic(instance) + instance._terminate = threading.Event() + try: + instance._threads = perf.thread_map() + evsel = perf.evsel(type = perf.TYPE_SOFTWARE, + config = perf.COUNT_SW_DUMMY, + task = 1, comm = 1, mmap = 0, freq = 0, + wakeup_events = 1, watermark = 1, + sample_type = perf.SAMPLE_TID | perf.SAMPLE_CPU) + evsel.open(cpus = self._cpus, threads = instance._threads) + instance._evlist = perf.evlist(self._cpus, instance._threads) + instance._evlist.add(evsel) + if instance._perf_mmap_pages is None: + instance._evlist.mmap() + else: + instance._evlist.mmap(pages = instance._perf_mmap_pages) + instance._thread = threading.Thread(target = self._thread_code, args = [instance]) + instance._thread.start() + # no perf + except Exception: + instance._dynamic_tuning_enabled = False + @classmethod def _get_config_options(cls): return { @@ -949,7 +951,7 @@ def _instance_apply_static(self, instance): and len(vals) == 5] sched_cfg = sorted(buf, key=lambda option_vals: option_vals[1][0]) sched_all = dict() - # for runtime tuning + # for dynamic tuning instance._sched_lookup = {} for option, (rule_prio, scheduler, priority, affinity, regex) \ in sched_cfg: @@ -973,9 +975,6 @@ def _instance_apply_static(self, instance): priority, affinity) self._storage.set(self._scheduler_storage_key, self._scheduler_original) - if self._daemon and instance._runtime_tuning: - instance._thread = threading.Thread(target = self._thread_code, args = [instance]) - instance._thread.start() def _restore_ps_affinity(self): try: @@ -1020,9 +1019,6 @@ def _cgroup_cleanup_tasks(self): def _instance_unapply_static(self, instance, rollback = consts.ROLLBACK_SOFT): super(SchedulerPlugin, self)._instance_unapply_static(instance, rollback) - if self._daemon and instance._runtime_tuning: - instance._terminate.set() - instance._thread.join() self._restore_ps_affinity() self._cgroup_restore_affinity() self._cgroup_cleanup_tasks() @@ -1031,6 +1027,10 @@ def _instance_unapply_static(self, instance, rollback = consts.ROLLBACK_SOFT): if self._cgroup_mount_point_init: self._cgroup_finalize() + def _instance_deinit_dynamic(self, instance): + instance._terminate.set() + instance._thread.join() + def _cgroup_verify_affinity_one(self, cgroup, affinity): log.debug("Verifying cgroup '%s' affinity" % cgroup) path = "%s/%s/%s" % (self._cgroup_mount_point, cgroup, "cpuset.cpus")