Initial mount timeout should be configurable

[nodakai]

juicefs/cmd/mount_unix.go

Line 43 in a612e08

for i := 0; i < 20; i++ {

I've encountered an issue with YugabyteDB, a PostgreSQL-compatible distributed database. It generally works well but has a quirk: its system tables respond slowly without human optimization, e.g. yugabyte/yugabyte-db#7745

Specifically, the 10-second timeout in JuiceFS is insufficient for XORM to retrieve table metadata from Yugabyte using this complex query with multiple joins. (I know XORM has moved to GiTea, but this part hasn't changed much)

I anticipate that YugabyteDB will perform adequately once it overcomes the initial table metadata retrieval hurdle. Could you introduce a new parameter or environment variable to extend the hard-coded 20 x 0.5s = 10s timeout, allowing for longer durations as needed?

[davies]

We could add an environment variable as a temporary workaround, which could be JFS_MOUNT_TIMEOUT=10.