upgrade MongoShake to version >= 2.4.0] error on V2.4.1

[k-nuhdim]

Below are snippet of collecter.log

[2020/04/18 13:39:55 IST] [CRIT] [collector.(*OplogSyncer).Start:194] load checkpoint[testdb] failed[current required checkpoint version[1] > input[0], please upgrade MongoShake to version >= 2.4.0]

Please help me on this.

[vinllen]

You should remove the old checkpoint which locates on the source MongoDB by default, 'cause the checkpoint format is changed since v2.4.0

[k-nuhdim]

Deleted all monoshake dbs from source database , but still egtting same issue, does that saving checkpoint anywhere else?

[vinllen]

if the source MongoDB is replica, the address is mongoshake.ckpt_default on the source MongoDB. if the source MongoDB is sharding, the address is admin.ckpt_default on config-server. Once delete the checkpoint, if you start with sync_mode = all, MongoShake will run full-sync and then incremental sync; if you start with sync_mode = oplog, MongoShake will run incremental sync only to fetch the oldest oplog from souce MongoDB.

[k-nuhdim]

Hi vinellen, Just want to know how the " sync_mode = all " works , is it useful to sync all data from one replicaset to another ?

[vinllen]

If you want to sync all data from one replica to another, please set sync_mode = all. There're two stages of syncing:

1. full sync. Scan all data on the source replica/sharding and then write to the target mongodb.
2. incremental sync. Fetch the oplog from the source and then apply it to the target.

So sync_mode = all means full+incremental, sync_mode = full means only full sync, sync_mode = incr means only incremental sync.

[k-nuhdim]

@vinllen - Thanks ! my use case is to replicate Production server data -----> DR Server and once its fail-over to DR happen then I need to change syncing direction from DR ----> PROD . so i hope if i make sync mode as full + increment it will sync all missing data first the later oplog will replaying based on the source write.

[vinllen]

Currently, MongoShake only supports failover in the incremental stage which means if you restart MongoShake or MongoDB crash in full sync stage, you need to restart MongoShake to run full+incrmental sync again. But in the incremental stage, MongoShake supports resume-from-break-point.

[k-nuhdim]

@vinllen - Thanks for the confirmation - Closing this.