Open pyromaniac3010 opened 2 years ago
After a full sync you'll need to bring down the Docker stack and bring it back up again (removing the DB restore field).
Ok, so I executed
docker-compose down
docker-compuse up -d
Now the graphgl-port does respond (after about 30 minute wait), but the database is now reverted to the last database block and is many hours back again. So I have some questions / suggestions for improvement:
There should be a hint in the README or Wiki to tell everyone that you NEED to execute "docker-compose down && docker-compuse up -d", if you follow the suggested install path
Why do we have to restart the whole stack? It would be great if a "docker-compose restart cardano-graphql" would work, but it did not. This should prevent the rollback of the cardano-node / postgres db states. There are no listed dependencys in the docker-compose.yml which indicate that other services depend on this service or that cardano-graphql depends on other services.
Update System Requirements The system Requirements in https://github.com/input-output-hk/cardano-graphql/releases/tag/6.1.0 are outdated. There should be some information in the Readme / Wiki telling that you need REALLY FAST disks with LOTS of IOPS. Otherwise everything takes hours or days to startup / restart. And 80GB of diskspace is far from enough. The created docker-volumes for a current up to date setup are way bigger:
110G ./cardano-graphql_postgres-data
23G ./cardano-graphql_node-db
0 ./cardano-graphql_node-ipc
5.0G ./cardano-graphql_db-sync-data
Plus you need about 40GB - 50GB extra to restore the database snapshot. So you need at least 200GB hard disk space for the complete cardano-graphql docker stack.
Environment
The CLI can be used to gather generic information about your system to assist with support. No installation is required thanks to npx
Alternatively, what release or Git revision are you using:
Platform
"Amazon Linux 2"
Runtime
Runtime version:
Steps to reproduce the bug
I installed cardano-graphql 6.1.0 via docker-compose stack and RESTORE_SNAPSHOT=https://update-cardano-mainnet.iohk.io/cardano-db-sync/12/db-sync-snapshot-schema-12-block-6590499-x86_64.tgz
After starting the stack and wating 3 days, the cardano-node and cardano-db-sync-extended are fully synched (checked via postgres-query "select now () - max (time) as behind_by from block ; ". So I tried to get graphql responses via curl as stated in the docs. But I always get Connection reset by peer:
All containers are up and running:
The only errors I see in docker logs are coming from postgres-container:
What is the expected behavior?
cardano-graphql should answer correctly on graphql port 3100 and return a correct json string.