Real contribution to real open source

    A recent post about how we in Sberbank Technologies are developing Open Source has revealed many interesting details about approaches, aspirations and ideologies.

    Today I want to talk about the real contribution that our team makes to Open Source.

    The main activity of our team is the development of Apache Ignite and lately we have done a lot of useful things.

    So, then just the facts.

    1) In version 2.5, a new topology version appeared: the star around ZooKeeper .
    We participated in the stabilization of the new functionality and solved many problems.
    Now clients with a huge topology can not worry about the reliability of their clusters.

    2) jsr107 was updated to version 1.1 and we added support for it to the product.

    3) We finalized the rebalancing of the cluster so that the arrival and departure of client nodes does not affect it.

    4) Wefinalized IgniteSet, now the number of copies of distributed data structures does not affect the speed of individual instances.

    5) We added support for system views (SQL system view) for cluster diagnostics.
    Now it is easier for customers who are accustomed to administering RDBMS to get used to NoSQL grid.
    In this direction, work has just begun, and new views will continue to be added.

    6) We have banned transactions on atomic caches in transactions, since they could lead to inconsistency of data in the cluster in case of a rollback of the transaction and other surprises in data visibility.
    This allowed us to find a lot of errors in the tests and production code.

    7) Weadded a single mechanism for handling critical errors.
    Now the error in the critical worker will be processed according to the selected policy and will not go unnoticed.

    This is not a complete list of the problems solved lately, but the others are either too simple, or, on the contrary, will require writing a separate article to clarify the essence of the problem solved.

    For example , the problem of cluster hanging on a topology change in case of extremely rare events was solved here .

    And by the way, our team is ready to accept new members in MSK and SPB .

    Also popular now: