||The JSESSIONID cookie does not set the HttpOnly flag for Resource Manager or Analytics.
||Performance ETL: Slow device and component key lookups. New configuration option
max-key-fetch-batch-size (default 1000) controls the number of keys
retrieved at once.
||Performance ETL: Performance config does not scale.
||Device Availability report is loading only 0 and 100% availability.
||Nth Percentile data not aggregating - silent failure
||Provide visibility into when Analytics actually begins to
extract perf data. Added new batch states CONFIGURING and
READY. Removed batch state
||Do not extract user passwords.
||Analytics server: 5.x PERF extraction was serialized by metric.
Now extracts multiple metrics concurrently. The
perf_metric_processing_threads setting in the
met_table limits the number of concurrent requests made on the
||Performance ETL: Performance config extract overwhelmed zenhub.
Now all ETL services are located in their own service folder in Control Center
named PEFORMANCE (5.x only). There are no zenperfetl services running
in the collector service folders (5.x only). The original PERFORMANCE extractors
registered in Analytics need to be manually removed using the
remove_extractor.sh script in the
ZenossAnalyticsMonitoring ZenPack. The names of the extractors to
be removed will have the form of <collector-name>/<#>.
||Performance ETL: One of many zenperfetl is slow
||Analytics server: Sub-optimal v2 query API usage
||Analytics server: Optimize the timestamp formatting in perf
data extraction process
||Performance ETL: When a device is added, the perf batch is blocked until the
config cache is refreshed.
||Analytics server: Perf data extraction cannot extract
performance data for devices with a very high number of components (tens of
thousands). Implemented the following new settings.
- query_tag_threshold (default 1000). Use wildcard queries unless
the number of time series results for a metric exceeds this value. If exceeded,
specify the tags for specific time series using multiple queries instead.
- query_max_tags_per_metric (default 200). When specifying tags
to select time series, the number of tags specified in one query shall not exceed
- query_min_tags_per_metric (default 10). When tags are used to
select time series, queries will not be subdivided if the number of tags specified
in one query falls below this value.
||Optimize the meta_etl clean up query in partition rotation
||dim_zenoss_instance not updated when config is saved.
||Limited visibility on batch processings. Have now added structured logging and
improved the INFO and DEBUG log output.
||Cisco UCS Fabric Interconnects Analytics aliases for
mem_pct produces negative values in Analytics.
||Two aliases defined for cpu__pct in one template.
||zenmodeletl chronically slow due to hits to query service every
||Alias management: ZenETL: Remove all ZenPack aliases from
||Need a dimension table representing time for Hourly data queries in Analytics.
||Alias management: Rewrite process for alias management. Now have new scripts
dumpAliases.py and manageAliases.py and a new