/opt/etcd/etcdctl --user=root:'123456' --endpoints=http://127.0.0.1:2379 defrag

[b03jsj]

{"level":"warn","ts":"2023-04-07T17:27:11.380+0800","logger":"etcd-client","caller":"v3/retry_interceptor.go:62","msg":"retrying of unary invoker failed","target":"etcd-endpoints://0xc0003d8a80/127.0.0.1:2379","attempt":0,"error":"rpc error: code = DeadlineExceeded desc = context deadline exceeded"}
Failed to defragment etcd member[http://127.0.0.1:2379] (context deadline exceeded)

dbsize 2.5 GB

why execute defragment got error like "context deadline exceeded"

if dbsize level is low , then defragment is ok. if dbsize high, then failed

and last ，the dbsize not change

this is my etcd config param:

name: 'etcd01'
data-dir: '/opt/etcd/etcd-data'
listen-client-urls: 'http://0.0.0.0:2379'
listen-peer-urls: 'http://0.0.0.0:2380'
advertise-client-urls: 'http://0.0.0.0:2379'
enable-v2: true
quota-backend-bytes: 7589934592
auto-compaction-mode: 'revision'
auto-compaction-retention: '10'

[chaochn47]

why execute defragment got error like "context deadline exceeded"

if dbsize level is low , then defragment is ok. if dbsize high, then failed

Hi @b03jsj, from my past experience, the error context deadline exceeded is because defragment handling in etcd server could take longer than the flag --command-timeout value of etcdctl, (which is by default 5s), especially for large db, then grpc based etcd client will give up and throw context deadline exceeded error instead of waiting indefinitely.

Could you please adjust the --command-timeout for longer time? Maybe starting with 1m.

[b03jsj]

yes，exactly as you say，defragment is running in background，continue about 1min，and successed，but dbsize not change，why？

[chaochn47]

Defragment will re-write the db file and remove the free pages in the old db file. If there is no free pages in the old db file, it is expected that db size won't be changed.

You can refer to etcd_mvcc_db_total_size_in_bytes - etcd_mvcc_db_total_size_in_use_in_bytes to estimate the fragmentation bytes by curl -sL <etcd-endpoints>/metrics | grep etcd_mvcc_db_total_size manually.

[b03jsj]

thanks！

[vwalek]

You might want to try compacting the database first and then defrag.
https://etcd.io/docs/v3.5/op-guide/maintenance/#history-compaction-v3-api-key-value-database

[b03jsj]

thanks for help
compact is already done，defrag is next step
I check curl -sL /metrics | grep etcd_mvcc_db_total_size，there is no free pages in the old db file，just like chaochn47 say

thanks all