You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@celeborn.apache.org by et...@apache.org on 2024/01/16 08:20:22 UTC
(incubator-celeborn) branch branch-0.4 updated: [CELEBORN-1223] Align master and worker metrics of document with MasterSource and WorkerSource
This is an automated email from the ASF dual-hosted git repository.
ethanfeng pushed a commit to branch branch-0.4
in repository https://gitbox.apache.org/repos/asf/incubator-celeborn.git
The following commit(s) were added to refs/heads/branch-0.4 by this push:
new 576f0cfac [CELEBORN-1223] Align master and worker metrics of document with MasterSource and WorkerSource
576f0cfac is described below
commit 576f0cfac569fb29b2a911d8d28f36266c2e9a11
Author: SteNicholas <pr...@163.com>
AuthorDate: Tue Jan 16 16:19:39 2024 +0800
[CELEBORN-1223] Align master and worker metrics of document with MasterSource and WorkerSource
### What changes were proposed in this pull request?
Align master and worker metrics of document with `MasterSource` and `WorkerSource` in `METRICS.md` and `monitoring.md`.
### Why are the changes needed?
Metrics of master and worker is inconsistent with `MasterSource` and `WorkerSource` at present. It is recommended to align master and worker metrics of document with `MasterSource` and `WorkerSource`:
- PushDataHandshakeFailCount
- RegionStartFailCount
- RegionFinishFailCount
- PrimaryPushDataHandshakeTime
- ReplicaPushDataHandshakeTime
- PrimaryRegionStartTime
- ReplicaRegionStartTime
- PrimaryRegionFinishTime
- ReplicaRegionFinishTime
- ActiveConnectionCount
- BufferStreamReadBuffer
- ReadBufferDispatcherRequestsLength
- ReadBufferAllocatedCount
- CreditStreamCount
- ActiveMapPartitionCount
- DeviceOSFreeBytes
- DeviceOSTotalBytes
- DeviceCelebornFreeBytes
- DeviceCelebornTotalBytes
- PotentialConsumeSpeed
- UserProduceSpeed
- WorkerConsumeSpeed
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
No.
Closes #2226 from SteNicholas/CELEBORN-1223.
Authored-by: SteNicholas <pr...@163.com>
Signed-off-by: mingji <fe...@alibaba-inc.com>
(cherry picked from commit 402d23d0ea6d8bdcad412322610b6b7d41013812)
Signed-off-by: mingji <fe...@alibaba-inc.com>
---
METRICS.md | 196 ++++++++++++++++++++++++++++++-----------------------
docs/monitoring.md | 51 +++++++-------
2 files changed, 136 insertions(+), 111 deletions(-)
diff --git a/METRICS.md b/METRICS.md
index 781390ee9..db54580c5 100644
--- a/METRICS.md
+++ b/METRICS.md
@@ -74,92 +74,116 @@ Here is an example of Grafana dashboard importing.
## Details
-| MetricName | Scope | Description |
-|:--------------------------------------:|:-----------------:|:---------------------------------------------------------------------------------------------------------------:|
-| WorkerCount | master | The count of active workers. |
-| ExcludedWorkerCount | master | The count of workers in excluded list. |
-| OfferSlotsTime | master | The time of offer slots. |
-| PartitionSize | master | The estimated partition size of last 20 flush window whose length is 15 seconds by defaults. |
-| RegisteredShuffleCount | master and worker | The value means count of registered shuffle. |
-| RunningApplicationCount | master and worker | The value means count of running applications. |
-| ActiveShuffleSize | master and worker | The value means the active shuffle size for workers or a worker including master replica and slave replica. |
-| ActiveShuffleFileCount | master and worker | The value means the active shuffle size for workers or a worker including master replica and slave replica. |
-| diskFileCount | master and worker | The count of disk files consumption by each user. |
-| diskBytesWritten | master and worker | The amount of disk files consumption by each user. |
-| hdfsFileCount | master and worker | The count of hdfs files consumption by each user. |
-| hdfsBytesWritten | master and worker | The amount of hdfs files consumption by each user. |
-| CommitFilesTime | worker | CommitFiles means flush and close a shuffle partition file. |
-| ReserveSlotsTime | worker | ReserveSlots means acquire a disk buffer and record partition location. |
-| FlushDataTime | worker | FlushData means flush a disk buffer to disk. |
-| OpenStreamTime | worker | OpenStream means read a shuffle file and send client about chunks size and stream index. |
-| FetchChunkTime | worker | FetchChunk means read a chunk from a shuffle file and send to client. |
-| ChunkStreamCount | worker | The stream count for reduce partition reading streams in current worker. |
-| OpenStreamFailCount | worker | The count of opening stream failed in current worker. |
-| FetchChunkFailCount | worker | The count of fetching chunk failed in current worker. |
-| PrimaryPushDataTime | worker | PrimaryPushData means handle pushdata of primary partition location. |
-| ReplicaPushDataTime | worker | ReplicaPushData means handle pushdata of replica partition location. |
-| WriteDataHardSplitCount | worker | The count of writing PushData or PushMergedData to HARD_SPLIT partition in current worker. |
-| WriteDataFailCount | worker | The count of writing PushData or PushMergedData failed in current worker. |
-| ReplicateDataFailCount | worker | The count of replicating PushData or PushMergedData failed in current worker. |
-| ReplicateDataWriteFailCount | worker | The count of replicating PushData or PushMergedData failed caused by write failure in peer worker. |
-| ReplicateDataCreateConnectionFailCount | worker | The count of replicating PushData or PushMergedData failed caused by creating connection failed in peer worker. |
-| ReplicateDataConnectionExceptionCount | worker | The count of replicating PushData or PushMergedData failed caused by connection exception in peer worker. |
-| ReplicateDataTimeoutCount | worker | The count of replicating PushData or PushMergedData failed caused by push timeout in peer worker. |
-| TakeBufferTime | worker | TakeBuffer means get a disk buffer from disk flusher. |
-| SlotsAllocated | worker | Slots allocated in last hour |
-| NettyMemory | worker | The value measures all kinds of transport memory used by netty. |
-| SortTime | worker | SortTime measures the time used by sorting a shuffle file. |
-| SortMemory | worker | SortMemory means total reserved memory for sorting shuffle files . |
-| SortingFiles | worker | This value means the count of sorting shuffle files. |
-| SortedFiles | worker | This value means the count of sorted shuffle files. |
-| SortedFileSize | worker | This value means the count of sorted shuffle files 's total size. |
-| DiskBuffer | worker | Disk buffers are part of netty used memory, means data need to write to disk but haven't been written to disk. |
-| PausePushDataTime | worker | PausePushData means stop receiving data from client. |
-| PausePushDataAndReplicateTime | worker | PausePushDataAndReplicate means stop receiving data from client and other workers. |
-| PausePushData | worker | The count of stopping receiving data from client in current worker. |
-| PausePushDataAndReplicate | worker | The count of stopping receiving data from client and other workers in current worker. |
-| jvm_gc_count | JVM | The GC count of each garbage collector. |
-| jvm_gc_time | JVM | The GC cost time of each garbage collector. |
-| jvm_memory_heap_init | JVM | The amount of heap init memory. |
-| jvm_memory_heap_max | JVM | The amount of heap max memory. |
-| jvm_memory_heap_used | JVM | The amount of heap used memory. |
-| jvm_memory_heap_committed | JVM | The amount of heap committed memory. |
-| jvm_memory_heap_usage | JVM | The percentage of heap memory usage. |
-| jvm_memory_non_heap_init | JVM | The amount of non-heap init memory. |
-| jvm_memory_non_heap_max | JVM | The amount of non-heap max memory. |
-| jvm_memory_non_heap_used | JVM | The amount of non-heap uesd memory. |
-| jvm_memory_non_heap_committed | JVM | The amount of non-heap committed memory. |
-| jvm_memory_non_heap_usage | JVM | The percentage of non-heap memory usage. |
-| jvm_memory_pools_init | JVM | The amount of each memory pool's init memory. |
-| jvm_memory_pools_max | JVM | The amount of each memory pool's max memory. |
-| jvm_memory_pools_used | JVM | The amount of each memory pool's used memory. |
-| jvm_memory_pools_committed | JVM | The amount of each memory pool's committed memory. |
-| jvm_memory_pools_used_after_gc | JVM | The amount of each memory pool's used memory after GC. |
-| jvm_memory_pools_usage | JVM | The percentage of each memory pool's memory usage. |
-| jvm_memory_total_init | JVM | The amount of total init memory. |
-| jvm_memory_total_max | JVM | The amount of total max memory. |
-| jvm_memory_total_used | JVM | The amount of total used memory. |
-| jvm_memory_total_committed | JVM | The amount of each memory pool's committed memory. |
-| jvm_direct_capacity | JVM | An estimate of the total capacity of the buffers in this pool |
-| jvm_direct_count | JVM | An estimate of the number of buffers in the pool |
-| jvm_direct_used | JVM | An estimate of the memory that JVM is using for this buffer pool |
-| jvm_mapped_capacity | JVM | An estimate of the total capacity of the buffers in this pool |
-| jvm_mapped_count | JVM | An estimate of the number of buffers in the pool |
-| jvm_mapped_used | JVM | An estimate of the memory that JVM is using for this buffer pool |
-| jvm_thread_count | JVM | The current number of threads. |
-| jvm_thread_daemon_count | JVM | The current number of daemon threads. |
-| jvm_thread_blocked_count | JVM | The current number of threads having blocked state. |
-| jvm_thread_deadlock_count | JVM | The current number of threads having deadlock state. |
-| jvm_thread_new_count | JVM | The current number of threads having new state. |
-| jvm_thread_runnable_count | JVM | The current number of threads having runnable state. |
-| jvm_thread_terminated_count | JVM | The current number of threads having terminated state. |
-| jvm_thread_timed_waiting_count | JVM | The current number of threads having timed_waiting state. |
-| jvm_thread_waiting_count | JVM | The current number of threads having waiting state. |
-| jvm_classloader_loaded | JVM | The total number of classes loaded since the start of the JVM. |
-| jvm_classloader_unloaded | JVM | The total number of classes unloaded since the start of the JVM. |
-| JVMCPUTime | system | The JVM costs cpu time. |
-| AvailableProcessors | system | The amount of system available processors. |
-| LastMinuteSystemLoad | system | The last minute load of system. |
+| MetricName | Scope | Description |
+|:--------------------------------------:|:-----------------:|:-----------------------------------------------------------------------------------------------------------------:|
+| RegisteredShuffleCount | master and worker | The value means count of registered shuffle. |
+| RunningApplicationCount | master and worker | The value means count of running applications. |
+| ActiveShuffleSize | master and worker | The value means the active shuffle size for workers or a worker including master replica and slave replica. |
+| ActiveShuffleFileCount | master and worker | The value means the active shuffle file count for workers or a worker including master replica and slave replica. |
+| diskFileCount | master and worker | The count of disk files consumption by each user. |
+| diskBytesWritten | master and worker | The amount of disk files consumption by each user. |
+| hdfsFileCount | master and worker | The count of hdfs files consumption by each user. |
+| hdfsBytesWritten | master and worker | The amount of hdfs files consumption by each user. |
+| WorkerCount | master | The count of active workers. |
+| LostWorkers | master | The count of workers in lost list. |
+| ExcludedWorkerCount | master | The count of workers in excluded list. |
+| IsActiveMaster | master | Whether the current master is active. |
+| PartitionSize | master | The estimated partition size of last 20 flush window whose length is 15 seconds by defaults. |
+| OfferSlotsTime | master | The time of offer slots. |
+| OpenStreamTime | worker | OpenStream means read a shuffle file and send client about chunks size and stream index. |
+| FetchChunkTime | worker | FetchChunk means read a chunk from a shuffle file and send to client. |
+| ChunkStreamCount | worker | The stream count for reduce partition reading streams in current worker. |
+| OpenStreamFailCount | worker | The count of opening stream failed in current worker. |
+| FetchChunkFailCount | worker | The count of fetching chunk failed in current worker. |
+| PrimaryPushDataTime | worker | PrimaryPushData means handle PushData of primary partition location. |
+| ReplicaPushDataTime | worker | ReplicaPushData means handle PushData of replica partition location. |
+| WriteDataHardSplitCount | worker | The count of writing PushData or PushMergedData to HARD_SPLIT partition in current worker. |
+| WriteDataFailCount | worker | The count of writing PushData or PushMergedData failed in current worker. |
+| ReplicateDataFailCount | worker | The count of replicating PushData or PushMergedData failed in current worker. |
+| ReplicateDataWriteFailCount | worker | The count of replicating PushData or PushMergedData failed caused by write failure in peer worker. |
+| ReplicateDataCreateConnectionFailCount | worker | The count of replicating PushData or PushMergedData failed caused by creating connection failed in peer worker. |
+| ReplicateDataConnectionExceptionCount | worker | The count of replicating PushData or PushMergedData failed caused by connection exception in peer worker. |
+| ReplicateDataTimeoutCount | worker | The count of replicating PushData or PushMergedData failed caused by push timeout in peer worker. |
+| PushDataHandshakeFailCount | worker | The count of PushDataHandshake failed in current worker. |
+| RegionStartFailCount | worker | The count of RegionStart failed in current worker. |
+| RegionFinishFailCount | worker | The count of RegionFinish failed in current worker. |
+| PrimaryPushDataHandshakeTime | worker | PrimaryPushDataHandshake means handle PushData of primary partition location. |
+| ReplicaPushDataHandshakeTime | worker | ReplicaPushDataHandshake means handle PushData of replica partition location. |
+| PrimaryRegionStartTime | worker | PrimaryRegionStart means handle RegionStart of primary partition location. |
+| ReplicaRegionStartTime | worker | ReplicaRegionStart means handle RegionStart of replica partition location. |
+| PrimaryRegionFinishTime | worker | PrimaryRegionFinish means handle RegionFinish of primary partition location. |
+| ReplicaRegionFinishTime | worker | ReplicaRegionFinish means handle RegionFinish of replica partition location. |
+| PausePushDataTime | worker | PausePushData means stop receiving data from client. |
+| PausePushDataAndReplicateTime | worker | PausePushDataAndReplicate means stop receiving data from client and other workers. |
+| PausePushData | worker | The count of stopping receiving data from client in current worker. |
+| PausePushDataAndReplicate | worker | The count of stopping receiving data from client and other workers in current worker. |
+| TakeBufferTime | worker | TakeBuffer means get a disk buffer from disk flusher. |
+| FlushDataTime | worker | FlushData means flush a disk buffer to disk. |
+| CommitFilesTime | worker | CommitFiles means flush and close a shuffle partition file. |
+| SlotsAllocated | worker | Slots allocated in last hour |
+| ReserveSlotsTime | worker | ReserveSlots means acquire a disk buffer and record partition location. |
+| ActiveConnectionCount | worker | The value means count of active network connection. |
+| NettyMemory | worker | The value measures all kinds of transport memory used by netty. |
+| SortTime | worker | SortTime measures the time used by sorting a shuffle file. |
+| SortMemory | worker | SortMemory means total reserved memory for sorting shuffle files . |
+| SortingFiles | worker | This value means the count of sorting shuffle files. |
+| SortedFiles | worker | This value means the count of sorted shuffle files. |
+| SortedFileSize | worker | This value means the count of sorted shuffle files 's total size. |
+| DiskBuffer | worker | Disk buffers are part of netty used memory, means data need to write to disk but haven't been written to disk. |
+| BufferStreamReadBuffer | worker | This value means memory used by credit stream read buffer. |
+| ReadBufferDispatcherRequestsLength | worker | This value means the queue size of read buffer allocation requests. |
+| ReadBufferAllocatedCount | worker | This value means count of allocated read buffer. |
+| CreditStreamCount | worker | This value means count of stream for map partition reading streams. |
+| ActiveMapPartitionCount | worker | This value means count of active map partition reading streams. |
+| DeviceOSFreeBytes | worker | This value means actual usable space of OS for device monitor. |
+| DeviceOSTotalBytes | worker | This value means total usable space of OS for device monitor. |
+| DeviceCelebornFreeBytes | worker | This value means actual usable space of Celeborn for device monitor. |
+| DeviceCelebornTotalBytes | worker | This value means configured usable space of Celeborn for device monitor. |
+| PotentialConsumeSpeed | worker | This value means speed of potential consumption for congestion control. |
+| UserProduceSpeed | worker | This value means speed of user production for congestion control. |
+| WorkerConsumeSpeed | worker | This value means speed of worker consumption for congestion control. |
+| jvm_gc_count | JVM | The GC count of each garbage collector. |
+| jvm_gc_time | JVM | The GC cost time of each garbage collector. |
+| jvm_memory_heap_init | JVM | The amount of heap init memory. |
+| jvm_memory_heap_max | JVM | The amount of heap max memory. |
+| jvm_memory_heap_used | JVM | The amount of heap used memory. |
+| jvm_memory_heap_committed | JVM | The amount of heap committed memory. |
+| jvm_memory_heap_usage | JVM | The percentage of heap memory usage. |
+| jvm_memory_non_heap_init | JVM | The amount of non-heap init memory. |
+| jvm_memory_non_heap_max | JVM | The amount of non-heap max memory. |
+| jvm_memory_non_heap_used | JVM | The amount of non-heap uesd memory. |
+| jvm_memory_non_heap_committed | JVM | The amount of non-heap committed memory. |
+| jvm_memory_non_heap_usage | JVM | The percentage of non-heap memory usage. |
+| jvm_memory_pools_init | JVM | The amount of each memory pool's init memory. |
+| jvm_memory_pools_max | JVM | The amount of each memory pool's max memory. |
+| jvm_memory_pools_used | JVM | The amount of each memory pool's used memory. |
+| jvm_memory_pools_committed | JVM | The amount of each memory pool's committed memory. |
+| jvm_memory_pools_used_after_gc | JVM | The amount of each memory pool's used memory after GC. |
+| jvm_memory_pools_usage | JVM | The percentage of each memory pool's memory usage. |
+| jvm_memory_total_init | JVM | The amount of total init memory. |
+| jvm_memory_total_max | JVM | The amount of total max memory. |
+| jvm_memory_total_used | JVM | The amount of total used memory. |
+| jvm_memory_total_committed | JVM | The amount of each memory pool's committed memory. |
+| jvm_direct_capacity | JVM | An estimate of the total capacity of the buffers in this pool |
+| jvm_direct_count | JVM | An estimate of the number of buffers in the pool |
+| jvm_direct_used | JVM | An estimate of the memory that JVM is using for this buffer pool |
+| jvm_mapped_capacity | JVM | An estimate of the total capacity of the buffers in this pool |
+| jvm_mapped_count | JVM | An estimate of the number of buffers in the pool |
+| jvm_mapped_used | JVM | An estimate of the memory that JVM is using for this buffer pool |
+| jvm_thread_count | JVM | The current number of threads. |
+| jvm_thread_daemon_count | JVM | The current number of daemon threads. |
+| jvm_thread_blocked_count | JVM | The current number of threads having blocked state. |
+| jvm_thread_deadlock_count | JVM | The current number of threads having deadlock state. |
+| jvm_thread_new_count | JVM | The current number of threads having new state. |
+| jvm_thread_runnable_count | JVM | The current number of threads having runnable state. |
+| jvm_thread_terminated_count | JVM | The current number of threads having terminated state. |
+| jvm_thread_timed_waiting_count | JVM | The current number of threads having timed_waiting state. |
+| jvm_thread_waiting_count | JVM | The current number of threads having waiting state. |
+| jvm_classloader_loaded | JVM | The total number of classes loaded since the start of the JVM. |
+| jvm_classloader_unloaded | JVM | The total number of classes unloaded since the start of the JVM. |
+| JVMCPUTime | system | The JVM costs cpu time. |
+| AvailableProcessors | system | The amount of system available processors. |
+| LastMinuteSystemLoad | system | The last minute load of system. |
## Implementation
diff --git a/docs/monitoring.md b/docs/monitoring.md
index b172f496e..73515f150 100644
--- a/docs/monitoring.md
+++ b/docs/monitoring.md
@@ -91,18 +91,18 @@ reported in the list.
These metrics are exposed by Celeborn master.
- namespace=master
- - WorkerCount
- - LostWorkers
- - ExcludedWorkerCount
- RegisteredShuffleCount
- RunningApplicationCount
- - IsActiveMaster
- - PartitionSize
- - The size of estimated shuffle partition.
- ActiveShuffleSize
- The active shuffle size of workers.
- ActiveShuffleFileCount
- The active shuffle file count of workers.
+ - WorkerCount
+ - LostWorkers
+ - ExcludedWorkerCount
+ - IsActiveMaster
+ - PartitionSize
+ - The size of estimated shuffle partition.
- OfferSlotsTime
- The time for masters to handle `RequestSlots` request when registering shuffle.
@@ -131,11 +131,12 @@ These metrics are exposed by Celeborn master.
These metrics are exposed by Celeborn worker.
- namespace=worker
- - CommitFilesTime
- - The time for a worker to flush buffers and close files related to specified shuffle.
- - ReserveSlotsTime
- - FlushDataTime
- - The time for a worker to write a buffer which is 256KB by default to storage.
+ - RegisteredShuffleCount
+ - RunningApplicationCount
+ - ActiveShuffleSize
+ - The active shuffle size of a worker including master replica and slave replica.
+ - ActiveShuffleFileCount
+ - The active shuffle file count of a worker including master replica and slave replica.
- OpenStreamTime
- The time for a worker to process openStream RPC and return StreamHandle.
- FetchChunkTime
@@ -164,11 +165,23 @@ These metrics are exposed by Celeborn worker.
- ReplicaRegionStartTime
- PrimaryRegionFinishTime
- ReplicaRegionFinishTime
+ - PausePushDataTime
+ - The time for a worker to stop receiving pushData from clients because of back pressure.
+ - PausePushDataAndReplicateTime
+ - The time for a worker to stop receiving pushData from clients and other workers because of back pressure.
+ - PausePushData
+ - The count for a worker to stop receiving pushData from clients because of back pressure.
+ - PausePushDataAndReplicate
+ - The count for a worker to stop receiving pushData from clients and other workers because of back pressure.
- TakeBufferTime
- The time for a worker to take out a buffer from a disk flusher.
- - RegisteredShuffleCount
- - RunningApplicationCount
+ - FlushDataTime
+ - The time for a worker to write a buffer which is 256KB by default to storage.
+ - CommitFilesTime
+ - The time for a worker to flush buffers and close files related to specified shuffle.
- SlotsAllocated
+ - ReserveSlotsTime
+ - ActiveConnectionCount
- NettyMemory
- The total amount of off-heap memory used by celeborn worker.
- SortTime
@@ -180,14 +193,6 @@ These metrics are exposed by Celeborn worker.
- SortedFileSize
- DiskBuffer
- The memory occupied by pushData and pushMergedData which should be written to disk.
- - PausePushDataTime
- - The time for a worker to stop receiving pushData from clients because of back pressure.
- - PausePushDataAndReplicateTime
- - The time for a worker to stop receiving pushData from clients and other workers because of back pressure.
- - PausePushData
- - The count for a worker to stop receiving pushData from clients because of back pressure.
- - PausePushDataAndReplicate
- - The count for a worker to stop receiving pushData from clients and other workers because of back pressure.
- BufferStreamReadBuffer
- The memory used by credit stream read buffer.
- ReadBufferDispatcherRequestsLength
@@ -204,10 +209,6 @@ These metrics are exposed by Celeborn worker.
- PotentialConsumeSpeed
- UserProduceSpeed
- WorkerConsumeSpeed
- - ActiveShuffleSize
- - The active shuffle size of a worker including master replica and slave replica.
- - ActiveShuffleFileCount
- - The active shuffle file count of a worker including master replica and slave replica.
- push_server_usedHeapMemory
- push_server_usedDirectMemory
- push_server_numAllocations