We have a product running with some issues because of two power off.
The H2 database is on 1.3.x version but we are intended to update something to improve resiliency and durability, specially diminishing database corruption chances because our client is upset with DB crashes and needs a solution in order to not buy a proprietary expensive license. It is a 24/7 governmental health system with 20 simultaneous transactions in the worst case. The database is 4GB and it is 4 months old, so it will get bigger. The server is windows 2012 and we could use the new resilient filesystem format. As the last option we would install a BSD with ZFS filesystem that is mature. We would update H2 to 1.4.x using MVStore=FALSE; MVCC=FALSE; LOCK_TIMEOUT=10000; LOCK_MODE=3; AUTOCOMMIT=ON; Delay_transaction=-1?? and trying to CHECKPOINT SYNC and /or call fsync() on every insert update delete operation. As the fsync() (if it works on Win2012) could slow write operations, using an SSD is not discarded. The JEE application is running on the same server, so it is best to run in embedded mode? We can use a second server with a cluster copy because we need to do some online backups, but I think we could do a software to "select ... offset" in an autonomous way. We need the most stable and reliable way to operate with no database crashes. At this file size, some HDD could take longer, so using SPLIT is an alternative. The server is 4 core xeon with 16GB RAM, so we could improve caches and use G1CC garbage collection. Another option is using a one connection queue pool to guarantee one transaction at once. There is some autovacuum option like postgreSQL? So, there are so many options that we need help to make the best choice. Thank you. -- You received this message because you are subscribed to the Google Groups "H2 Database" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/d/optout.
