[YARN-1529] Add Localization overhead metrics to NM

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.2.2, 2.10.1, 3.4.0, 3.3.1
• Component/s: nodemanager
• Assignee: Jim Brennan
• Reporter: Gera Shegalov
• Created: 23/Dec/13 22:44
• Updated: 10/Jun/21 08:15
• Resolved: 30/Jul/20 20:09

Description

Users are often unaware of localization cost that their jobs incur. To measure effectiveness of localization caches it is necessary to expose the overhead in the form of metrics.

We propose addition of the following metrics to NodeManagerMetrics.

When a container is about to launch, its set of LocalResources has to be fetched from a central location, typically on HDFS, that results in a number of download requests for the files missing in caches.

LocalizedFilesMissed: total files (requests) downloaded from DFS. Cache misses.

LocalizedFilesCached: total localization requests that were served from local caches. Cache hits.

LocalizedBytesMissed: total bytes downloaded from DFS due to cache misses.

LocalizedBytesCached: total bytes satisfied from local caches.

Localized(Files|Bytes)CachedRatio: percentage of localized (files|bytes) that were served out of cache: ratio = 100 * caches / (caches + misses)

LocalizationDownloadNanos: total elapsed time in nanoseconds for a container to go from ResourceRequestTransition to LocalizedTransition

Issue Links

• is related to: YARN-3363 add localization and container launch time to ContainerMetrics at NM to show these timing information for each active container.
• is required by: MAPREDUCE-5696 Add Localization counters to MR
• relates to: YARN-1492 truly shared cache for jars (jobjar/libjar)