Hello! Ignite Clients are called "Thick", and indeed, they are pretty heavy-weight. They are full functional nodes, and as such, they have to join topology when starting and properly leave topology when stopping. This is long process. It also has to wait for any pending transactions in cluster, that's where you are seeing unpredictable times.
Ignite Clients are fully thread safe. You should never need to have more than one client per long-running process. You should avoid having Ignite Clients in short-lived processes. Reuse Ignite Clients where possible. You can use JDBC Thin client, ODBC client or REST API (for simple use cases) to avoid said limitations. There's currently work underway to introduce more thin clients for various platforms. https://apacheignite-sql.readme.io/docs/jdbc-driver https://apacheignite.readme.io/docs/rest-api and other platforms & protocols. There's no hard limit on number of clients, but practically I would not recommend having more than a few dozens, and definitely not having clients come and go all the time. Regards, -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/