Meaning of seqno in grastate.dat?

[S. Guenther]

Hello,

according to the documentation the value -1 for seqno means, that the node has crashed.

We are running a 3 node cluster, where on node just failed completely.

The file grastate.dat on the other two node looks like this:

# GALERA saved state

version: 2.1

uuid:    b40c9b92-31e9-11e3-b8bd-261ce9c1898a

seqno:   -1

cert_index:

But according to the ouput of 'mysql -p -u root 'SHOW STATUS"' the two nodes are synced and error.log looks fine to me:

150122 13:08:03 [Note] WSREP: forgetting 1523f1b9-4d3b-11e3-b927-d75043f216ef (tcp://10.10.10.3:4567)

150122 13:08:03 [Note] WSREP: New COMPONENT: primary = yes, bootstrap = no, my_idx = 1, memb_num = 2

150122 13:08:03 [Note] WSREP: STATE EXCHANGE: Waiting for state UUID.

150122 13:08:03 [Note] WSREP: (605604cb-4d3b-11e3-a2e8-d7a21f286515, 'tcp://0.0.0.0:4567') turning message relay requesting off

150122 13:08:03 [Note] WSREP: STATE EXCHANGE: sent state msg: 514fcdd2-a22f-11e4-813c-5a1e5032a7e4

150122 13:08:03 [Note] WSREP: STATE EXCHANGE: got state msg: 514fcdd2-a22f-11e4-813c-5a1e5032a7e4 from 0 (node2)

150122 13:08:03 [Note] WSREP: STATE EXCHANGE: got state msg: 514fcdd2-a22f-11e4-813c-5a1e5032a7e4 from 1 (node3)

150122 13:08:03 [Note] WSREP: Quorum results:

        version    = 2,

        component  = PRIMARY,

        conf_id    = 52,

        members    = 2/2 (joined/total),

        act_id     = 37929,

        last_appl. = 37908,

        protocols  = 0/4/2 (gcs/repl/appl),

        group UUID = b40c9b92-31e9-11e3-b8bd-261ce9c1898a

150122 13:08:03 [Note] WSREP: Flow-control interval: [23, 23]

150122 13:08:03 [Note] WSREP: New cluster view: global state: b40c9b92-31e9-11e3-b8bd-261ce9c1898a:37929, view# 53: Primary, number of nodes: 2, my index: 1, protocol version 2

150122 13:08:03 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

150122 13:08:03 [Note] WSREP: Assign initial position for certification: 37929, protocol version: 2

150122 13:08:09 [Note] WSREP:  cleaning up 1523f1b9-4d3b-11e3-b927-d75043f216ef (tcp://10.10.10.3:4567)

So, should I worry about the "-1"?

Stefan

[Matthias Rieber]

Hello,

On Thu, 22 Jan 2015, S. Guenther wrote:

> Hello,
>
> according to the documentation the value -1 for seqno means, that the node
> has crashed.
>
> We are running a 3 node cluster, where on node just failed completely.
>
> The file grastate.dat on the other two node looks like this:
>
>
> # GALERA saved state
> version: 2.1
> uuid: b40c9b92-31e9-11e3-b8bd-261ce9c1898a
> seqno: -1
> cert_index:

Galera set seqno to -1 while it's running. During a clean shutdown it will
be set to the correct value to join the cluster again. When the node
crashes, the value will stay -1 to indicate that there wasn't a clean
shutdown.

Matthias

[S. Guenther]

Hello Matthias,

Galera set seqno to -1 while it's running. During a clean shutdown it will
be set to the correct value to join the cluster again. When the node
crashes, the value will stay -1 to indicate that there wasn't a clean
shutdown.

Matthias

good news!  ;-)

Thank you,

Stefan