I think railway doesn’t let you raise the host vm.max_map_count (it’s a kernel setting), so memgraph won’t run with a mounted volume there , you need vm.max_map_count>=262144.

options you have: run memgraph on a VPS/VM or k8s where you can sysctl -w vm.max_map_count=262144, use memgraph cloud/another managed graph db, or as a temporary hack run without

mounting /var/lib/memgraph (in-memory only , data lost on restart)

I also got into this conclusion, but was thinking of way to not migrate for now and just save data

so mounted volumes + memgraph persistence won’t work. quick workable hack to avoid migrating now:

1. run memgraph without mounting /var/lib/memgraph (ephemeral)

2. periodically export while it’s up:

   • using mgconsole: dump database;orcreate snapshot;

   • or via mage: call export_util.cypher_all('/tmp/export.cyp'); or call export_util.json('/tmp/export.json');

3. immediately upload the dump/snapshot off-instance (aws s3, http endpoint, git, etc.) , automate this with a tiny script/cron in the container or an external job that calls the export endpoint and pushes to s3

4. to restore: start memgraph elsewhere (or a new instance) and run the cypher file or load the snapshot

those are short term backup-only solution , for real persistence move to a host where you can sysctl -w vm.max_map_count=262144 or use memgraph cloud

feel free to reach out if you need further clarification

Ok