dump command problems

[timfeirg]

juicefs version 1.2.2+2024-12-02.92e1db2

• dump command did not finish, even after dumped inodes exceed file system usage
• --fast option breaks database connection (without --fast, dump works normally)

2024/12/30 05:24:45.086469 juicefs[31691] <INFO>: Meta address: mysql://xxx:***@xxx:3306)/juicefs
Snapshot keys: 0/9920492 [-----------] 0.0/s ETA: 0s
2024/12/30 05:24:50.492201 juicefs[31691] <FATAL>: Fetoch all metadata from DB: invalid connection [main.go:31]

this error is accompanied by MySQL errors as well:

2024-12-30T05:06:02.293563-08:00 760724 [Note] [MY-010914] [Server] Aborted connection 760724 to db: 'xxx' user: 'xxx' host: '[xxx](http://xxx/)' (Got an error reading communication packets).

• ETA is mostly meaningless information as it's always "0s", in reality it's going to run for hours, consider using file system used inodes as denominator to get a more useful estimation

[anysql]

This is due to incorrect user of database connection, a coding bug I think.

If a connection is reading the result from mysql server, this connection cannot be used to run another query, else the packet sequence is broken and the connection is broken.

[zhijian-pro]

1. invalid connection is because the us3 database restarted due to slow queries, causing all sessions to be closed.
2. We will use total inodes to set the total number of the progress bar

[timfeirg]

dump command did not finish, even after dumped inodes exceed file system usage

don't forget to investigate this as well