Re: any way to monitor in real time the MaxBufferedEpoch value

[Maitrayi Sabaratnam]

Ndb nodes (more precisely the SUbscription MAnager block SUMA with id 257 at an Ndb node), send the requested data (epochs)to subscribers and keep them until the subscriber ack's. When acks get delayed, the epochs accumulate at SUMA. When the number of buffered epochs exceed the configured variable 'MaxBufferedEpochs' (default 100), Suma disconnects the subscriber.

There are 2 examples given below on how to retrieve the info on the unacc'd epochs piled at SUMA (via sql and ndbapi) from ndbinfo.ndb$pools table

1) using mysql:
'Used' shows the unacc'd epochs at an ndb node and 'total' shows the configured 'MaxBufferedEpochs'.

mysql>select node_id, pool_name, used, total from ndbinfo.ndb$pools where block_number=257 and pool_name='GCP';

mysql> select node_id, pool_name, used, total from ndbinfo.ndb$pools where block_number=257 and pool_name='GCP';
+---------+-----------+------+-------+
| node_id | pool_name | used | total |
+---------+-----------+------+-------+
| 2 | GCP | 53 | 100 |
| 3 | GCP | 53 | 100 |
+---------+-----------+------+-------+
2 rows in set (0.02 sec)

2) using ndbinfo_select_all tool:

storage/ndb/tools> ./ndbinfo_select_all -c <connectionstring> pools | grep "Subscription Record"
2 255 0 Subscription Record 0 100 100 144 0 0 0 0
3 255 0 Subscription Record 0 100 100 144 0 0 0 0