Is there a memory analysis tool for continuous TiKV memory growth?

This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: tikv内存持续增长,有没有内存分析工具

| username: 超7成网友

Old version v.3.0
Deployment: Mixed deployment, some servers have 2 TiKV, 1 TiDB, and 1 PD simultaneously, with 512G memory per server.
Issue: TiKV memory keeps increasing, block-cache is limited to 80G and has reached its limit. Currently, some TiKV memory usage has hit 200G.
I understand that RocksDB uses block cache for data caching and should release memory using LRU, but it doesn’t seem to be releasing memory.
Could you please advise on the possible reasons? Is there a memory analysis tool for TiKV? I would like to use a tool to analyze where the memory is being used.

| username: 小龙虾爱大龙虾 | Original post link

The V3 version does not have a direct memory analysis tool; higher versions of the dashboard can perform the analysis.

| username: 濱崎悟空 | Original post link

You can check the TiDB dashboard and monitoring.

| username: TiDBer_H5NdJb5Q | Original post link

I’m just afraid of these old systems, I don’t dare to touch them or upgrade them.

| username: Kongdom | Original post link

If the server has a UPS and the RAID card has a battery, you can consider setting this parameter to false for asynchronous disk writes. When our hardware was relatively poor in the early days, we set this parameter and saw significant improvements. However, in the event of an unexpected power outage, data loss can indeed occur, so use it with caution.

| username: zhaokede | Original post link

The version is quite old. If possible, it’s better to upgrade to a new version.

| username: zhh_912 | Original post link

Just adjust the configuration parameters of TiKV and RocksDB.

| username: WalterWj | Original post link

Take advantage of the current upgrade promotion to upgrade.

| username: tidb菜鸟一只 | Original post link

Try setting this a bit smaller, for example, 50G and see.