My SeaweedFS Filer instance connects to the SeaweedFS Master instance via a nat. I believe this is causing it to attempt to replicate from itself.
I'm happy with just having a single filer and handling HA at the cluster & store level, but is there a way I can completely disable the replication & peer discovery behaviour of the filer?
Logs attached if relevant:
W1228 23:12:39.414593 filer_server.go:145 skipping default store dir in ./filerldb2
I1228 23:12:39.417785 filer.go:164 create filer.store.id to -171668174
I1228 23:12:39.417926 configuration.go:28 configured filer store to mysql
I1228 23:12:39.419270 masterclient.go:210 master 10.5.3.3:9333 redirected to leader 10.5.3.3:9333
.I1228 23:12:39.595872 master_client.go:20 the cluster has 1 filer
I1228 23:12:39.596035 filer_server.go:172 10.5.3.3:8888 bootstrap from peers [node_type:"filer" address:"10.5.3.3:8888" is_add:true created_at_ns:1703805159421028874]
I1228 23:12:39.596167 filer.go:122 10.5.3.3:8888 aggregate from peers [10.5.3.3:8888]
I1228 23:12:39.598485 meta_aggregator.go:92 loopSubscribeToOneFiler read 10.5.3.3:8888 start from 2023-12-28 23:11:39.596026164 +0000 UTC 1703805099596026164
I1228 23:12:39.599431 filer.go:282 Start Seaweed Filer 30GB 3.59 27b34f379 at 10.5.3.3:8888
I1228 23:12:39.599442 meta_aggregator.go:103 subscribing remote 10.5.3.3:8888 meta change: connecting to peer filer 10.5.3.3:8888: rpc error: code = Unavailable desc = write tcp 172.17.0.5:58760->10.5.3.3:18888: write: broken pipe
I1228 23:12:41.332971 meta_aggregator.go:92 loopSubscribeToOneFiler read 10.5.3.3:8888 start from 2023-12-28 23:11:39.596026164 +0000 UTC 1703805099596026164
I1228 23:12:41.334287 meta_aggregator.go:189 subscribing remote 10.5.3.3:8888 meta change: 2023-12-28 23:11:39.596026164 +0000 UTC, clientId:1043354529
I1228 23:12:41.335386 filer_grpc_server_sub_meta.go:296 + local listener filer:10.5.3.3:[email protected]:35482 clientId -1043354529 clientEpoch 1
I1228 23:12:41.335426 filer_grpc_server_sub_meta.go:117 + filer:10.5.3.3:[email protected]:35482 local subscribe / from 2023-12-28 23:11:39.596026164 +0000 UTC clientId:-1043354529
I1228 23:12:41.335448 filer_grpc_server_sub_meta.go:130 read on disk filer:10.5.3.3:[email protected]:35482 local subscribe / from 2023-12-28 23:11:39.596026164 +0000 UTC
I1228 23:12:41.335887 filer_grpc_server_sub_meta.go:149 read in memory filer:10.5.3.3:[email protected]:35482 local subscribe / from 2023-12-28 23:11:39.596026164 +0000 UTC
It is a distributed system and is designed to have multiple nodes. So there are discovery stage and waiting stage.