Difference between revisions of "LDMS 3.x Plugins"
(→Production Store Plugins) |
(→Experimental) |
||
Line 121: | Line 121: | ||
|- | |- | ||
| tsampler || ? | | tsampler || ? | ||
+ | |- | ||
+ | | store_sos || Scalable Object Store storage prototype | ||
|} | |} |
Revision as of 10:26, 16 March 2018
You can list the stores and samplers and usage hints for your LDMS 3.4 installation with (typically)
/usr/bin/ldms_list_plugins.sh
This will list the compiled and installed plugins and their options for use in the LDMS configuration language. This may include experimental plugins. Generally, production plugins will have a man page, for example to understand the meminfo plugin, try
man Plugin_meminfo
Unfortunately, the man pages may not be installed on your system or they may not yet be complete. If anything in the table below provokes a question, please send it to the ovis-help mailing list described elsewhere.
Contents
Production Sampler Plugins
Below is a summary of the plugins that can be considered production quality in release 3.4.4 for commodity Linux environments.
name | metric set content |
---|---|
edac | memory error checking from /sys/devices/system/edac metrics |
jobid | Currently running job id and user (requires loose cooperation from queuing system, e.g. slurm) |
lnet_stats | /proc/sys/lnet/stats metrics, particularly memory |
lustre2_client | Lustre client metrics |
meminfo | /proc/meminfo values |
procinterrupts | interrupt counters (very large datasets on many-core machines) |
procnetdev | /proc/net tcp interface device counters (excludes rdma traffic/errors) |
procnfs | nfs v3 client statistics (calls, bytes) |
procstat | /proc/stat counters (includes cpu tick) |
sysclassib | infiniband counters and rates (includes rdma traffic/errors) |
vmstat | /proc/vmstat counters |
Production Store Plugins
name | output format | notes |
---|---|---|
store_csv | CSV files | requires the size of metric sets of a given name (schema) be identical across all nodes. |
store_flatfile | 1-metric files | narrow CSV file per metric with timestamp and source columns. Tolerates conflicting schema definitions. |
strore_rabbitv3 | AMQP messages | feeds all metrics to single AMQP broker via librabbitmq 0.8. |
Vendor specific
Below is a summary of the plugins that can be considered production quality in release 3.4.4 for CLE6
aries_mmr/nic_mmr/rtr_mmr | Cray XC aries network counters |
cray_system_sampler | wide variety of Cray XC metric |
These also work on appropriate platforms
cray_power_sampler | ? |
kgnilnd | kernel-space sampler for network data |
msr_interlagos | AMD CPU event counters |
Testing/Tutorial
all_example | (tutorial/testing) |
array_example | arrays (tutorial/testing) |
clock | ticker (tutorial/testing) |
fptrans | floating point transmission test sets (testing) |
generic_sampler | (tutorial/testing) |
synthetic | generates synthetic waveform data sets (testing) |
Need TLC, but can work under some circumstances
procdiskstats | ? |
procsensors | temperatures sensor |
sampler_atasmart | SMART counters from devices |
lustre2_mds | |
lustre2_oss |
Retiring if champion not self-identified
hadoop | ? |
knc_sampler | knights corner |
knc_sampler_copy | ditto |
knc_sampler_derived | ditto |
Experimental
hfclock | high frequency clock |
perfevent | Process perf event counters |
power_sampler | Power API 1 |
rapl | PAPI counters |
spapi | PAPI counters |
switchx | infiniband switch port data |
test_sampler | ? |
timer_base | ? |
tsampler | ? |
store_sos | Scalable Object Store storage prototype |