Thanos compact component test summary (processing historical data)

Reference documents

Official documents ：Thanos - Highly available Prometheus setup with long term storage capabilities

Translation and notes ：https://zhangrongjie.blog.csdn.net/article/details/120976651

Summary of component functions

compactor Components only interact with object storage , It is the only component that has the permission to delete the stored data of the object . There are two main functions ： Compress block（ Will be multiple block Merge into one ）、 Downsampling （ Can be disabled ,5m/1h Sampled data ）. Data retention time can be set , Raw data 、5m/1h The retention time of downsampled data can be set separately .

The test method

Copy some historical data , Use compact Handle （ Disable downsampling ）

Compact Set the parameters as follows ：

--downsampling.disable \

--delete-delay=0 \

--wait \

--wait-interval=5m \

--retention.resolution-raw=180d \

Test one

Copy 2020-10-20 To 2020-10-30 common 36.55GB data 、596 individual block(2h), To deal with .

compact First, each 4 individual 2h Of block Merge into one 8h Of block, And mark the processed block For delete status . After the processing is completed, each 5 individual 8h Of block Merge into one 40h Of block......

The processing time is too long , Failed to wait until the end of data processing . By kill fall compact Process time ,bucket The size has been changed from 36.55GB Add to about 70GB. The processing process is shown in figure. ：

  In the figure 1、2、3、4 Express 4 individual block flow （ from 4 individual Prometheus Uploaded data ）.

Test 1 Summary

Compaction The procedure increases the size of the storage , If, as described in the document , The maximum compression level is 2 Zhou's flower , be compaction The process will make bucket Triple the size . And the data deletion operation is in compaction And downsampling before .

Compact Will download all pending block To local , Upload after processing block. Therefore, the local disk should have at least 2 times 2 Available space for weekly data volume .

Test two

Copy 2020-09-21 To 2020-09-23 common 1.73GB data 、20 individual block(2h), To deal with .

Compact After completing the merger block After operation （ Because of the small amount of data , Only merged into 8h The level is terminated ）, The expired historical data has been deleted correctly . Process diagram ：

  Test II summary

Compact Data retention is available .