gocql: Cannot connect to scylla (Cassandra 2.2.8)
What version of Cassandra are you using?
cqlsh 5.0.1 | Cassandra 2.2.8 | CQL spec 3.3.1 | Native protocol v4 | Scylla version 1.7.4-0.20170726.ff643e3
What version of Gocql are you using?
Revision ce5020aabae349deda77bc1d559ef2300d3c93c5
What did you do?
use the following to spin up a local dev scylla instance
docker run --name some-scylla -p 9042:9042 -d scylladb/scylla --broadcast-address 127.0.0.1 --broadcast-rpc-address 127.0.0.1
and then ran
package main
import (
"fmt"
"github.com/gocql/gocql"
)
func main() {
cluster := gocql.NewCluster("127.0.0.1")
_, err := cluster.CreateSession()
if err != nil {
panic(err)
}
fmt.Println("cassandra init done")
}
What did you expect to see?
Everything connect and initialize correctly.
What did you see instead?
Cassandra Configuration failed: no connections were made when creating the session
Describe your Cassandra cluster
- output of
nodetool status
[root@f779e8aea5c1 /]# nodetool status
Datacenter: datacenter1
=======================
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
-- Address Load Tokens Owns Host ID Rack
UN 127.0.0.1 130.21 KB 256 ? 35a9b2db-a040-47bc-90b8-cdc6a56a17a6 rack1
Note: Non-system keyspaces don't have the same replication settings, effective ownership information is meaningless
- output of
SELECT peer, rpc_address FROM system.peers
peer | rpc_address
------+-------------
- rebuild your application with the
gocql_debugtag and post the output
2017/10/05 16:38:18 gocql: Session.handleNodeUp: 127.0.0.1:9042
2017/10/05 16:38:20 unable to dial "172.17.0.2": dial tcp 172.17.0.2:9042: i/o timeout
2017/10/05 16:38:20 gocql: Session.handleNodeDown: 172.17.0.2:9042
2017/10/05 16:38:20 gocql: Session.handleNodeUp: 172.17.0.2:9042
2017/10/05 16:38:22 unable to dial "172.17.0.2": dial tcp 172.17.0.2:9042: i/o timeout
panic: no connections were made when creating the session
About this issue
- Original URL
- State: closed
- Created 7 years ago
- Comments: 19 (9 by maintainers)
I had the same issue and fixed it by creating a docker network and adding a virtual interface to the host. The solution does not need to override or user default rout (0.0.0.0) for listen, broadcast and broadcast-rpc.
#add a virtual interface (os x) sudo ifconfig en0 inet 192.18.0.2 netmask 255.255.255.0 alias
#windows ifconfig eth0:1 10.0.0.11 netmask 255.255.255.0 up
#create docker subnetwork docker network create --subnet=192.18.0.0/24 skynet
#docker run scylla (you can always specify another ip for listen-address which need to be mapped to an alias on the host)
docker run --net skynet --ip 192.18.0.2 --name $NAME -p 192.18.0.2:9042:9042 -p 192.18.0.2:9160:9160 -d scylladb/scylla --overprovisioned 1
Clients will connect to the new adresse 192.18.0.2 i/o localhost
For those wondering where does 172.x.x.x come from, at startup scylla daemon creates system.local table and register container ip-address as listen-address(same for braodcast and broadcast-rpc). Clients retrieve those routes using cql and configure rpc (thrift to use those routes). if clients is containerized no issue as routing is handling by docker. But if clients is not containerized (eg client running on host) then it dials 172.x.x.x which does not exist on the host. I had this issue using scylladb with hadoop-gremlin for a janusgraph instance. Running a SparkGraphComputer fails.
I was able to fix this by having Scylla bind to all ports internally. Not sure what that means for related parties, but I think its a mac os docker networking issue with the scylla instance.