This is an automated email from the ASF dual-hosted git repository.
luzhijing pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/master by this push:
new 5570b57e41 [typo](doc)fix invalid url (#18855)
5570b57e41 is described below
commit 5570b57e41dd6b0a8ffd04666abee267b0e5b97c
Author: superche <[email protected]>
AuthorDate: Fri Apr 21 15:58:46 2023 +0800
[typo](doc)fix invalid url (#18855)
Co-authored-by: hechao <[email protected]>
---
.../admin-manual/maint-monitor/memory-management/be-oom-analysis.md | 2 +-
.../admin-manual/maint-monitor/memory-management/be-oom-analysis.md | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git
a/docs/en/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
b/docs/en/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
index c203dab762..1994cefe48 100644
---
a/docs/en/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
+++
b/docs/en/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
@@ -76,7 +76,7 @@ Memory Tracker Summary:
6. `type=load` imports a lot of memory.
7. When the `type=global` memory is used for a long time, continue to check
the `type=global` detailed statistics in the second half of the `Memory Tracker
Summary` log. When DataPageCache, IndexPageCache, SegmentCache, ChunkAllocator,
LastestSuccessChannelCache, etc. use a lot of memory, refer to [BE
Configuration Item](../../../admin-manual/config/be-config.md) to consider
modifying the size of the cache; when Orphan memory usage is too large,
Continue the analysis as follows.
- - If the sum of the tracker statistics of `Parent Label=Orphan` only
accounts for a small part of the Orphan memory, it means that there is
currently a large amount of memory that has no accurate statistics, such as the
memory of the brpc process. At this time, you can consider using the heap
profile [Memory Tracker]( ../../../../community/developer-guide/debug-tool) to
further analyze memory locations.
+ - If the sum of the tracker statistics of `Parent Label=Orphan` only
accounts for a small part of the Orphan memory, it means that there is
currently a large amount of memory that has no accurate statistics, such as the
memory of the brpc process. At this time, you can consider using the heap
profile [Memory Tracker](
https://doris.apache.org/community/developer-guide/debug-tool) to further
analyze memory locations.
- If the tracker statistics of `Parent Label=Orphan` account for most of
Orphan’s memory, when `Label=TabletManager` uses a lot of memory, further check
the number of tablets in the cluster. If there are too many tablets, delete
them and they will not be used table or data; when `Label=StorageEngine` uses
too much memory, further check the number of segment files in the cluster, and
consider manually triggering compaction if the number of segment files is too
large;
8. If `be/log/be.INFO` does not print the `Memory Tracker Summary` log before
OOM, it means that BE did not detect the memory limit in time, observe Grafana
memory monitoring to confirm the memory growth trend of BE before OOM, if OOM
is reproducible, consider adding `memory_debug=true` in `be.conf`, after
restarting the cluster, the cluster memory statistics will be printed every
second, observe the last `Memory Tracker Summary` log before OOM, and continue
to step 3 for analysis;
diff --git
a/docs/zh-CN/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
b/docs/zh-CN/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
index 1d58f9d325..1cd64d2b6d 100644
---
a/docs/zh-CN/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
+++
b/docs/zh-CN/docs/admin-manual/maint-monitor/memory-management/be-oom-analysis.md
@@ -76,7 +76,7 @@ Memory Tracker Summary:
6. `type=load`导入内存使用多时。
7. `type=global`内存使用多时，继续查看`Memory Tracker
Summary`日志后半部分已经打出得`type=global`详细统计。当
DataPageCache、IndexPageCache、SegmentCache、ChunkAllocator、LastestSuccessChannelCache
等内存使用多时，参考 [BE 配置项](../../../admin-manual/config/be-config.md) 考虑修改cache的大小；当
Orphan 内存使用过多时，如下继续分析。
- - 若`Parent Label=Orphan`的tracker统计值相加只占 Orphan 内存的小部分，则说明当前有大量内存没有准确统计，比如
brpc 过程的内存，此时可以考虑借助 heap profile [Memory
Tracker](../../../../community/developer-guide/debug-tool.md) 中的方法进一步分析内存位置。
+ - 若`Parent Label=Orphan`的tracker统计值相加只占 Orphan 内存的小部分，则说明当前有大量内存没有准确统计，比如
brpc 过程的内存，此时可以考虑借助 heap profile [Memory
Tracker](https://doris.apache.org/zh-CN/community/developer-guide/debug-tool)
中的方法进一步分析内存位置。
- 若`Parent Label=Orphan`的tracker统计值相加占 Orphan
内存的大部分，当`Label=TabletManager`内存使用多时，进一步查看集群 Tablet 数量，若 Tablet
数量过多则考虑删除过时不会被使用的表或数据；当`Label=StorageEngine`内存使用过多时，进一步查看集群 Segment 文件个数，若
Segment 文件个数过多则考虑手动触发compaction；
8. 若`be/log/be.INFO`没有在 OOM 前打印出`Memory Tracker Summary`日志，说明 BE
没有及时检测出内存超限，观察 Grafana 内存监控确认BE在 OOM 前的内存增长趋势，若 OOM
可复现，考虑在`be.conf`中增加`memory_debug=true`，重启集群后会每秒打印集群内存统计，观察 OOM 前的最后一次`Memory
Tracker Summary`日志，继续步骤3分析；
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
