The following are the highlights in this release:
This bug occurred a few times in a week in our production. Each time causes one server restart. For user who experiences this bug we suggest you to upgrade to 1.11.6.
Related PR: XiaoMi/rdsn#272, XiaoMi/rdsn#279, XiaoMi/rdsn#277
Related Issue: #315
Server connection threshold support: XiaoMi/rdsn#234
Whitelist support: XiaoMi/rdsn#228, XiaoMi/rdsn#226
Related Docs: https://pegasus.apache.org/administration/whitelist
In XiaoMi, some of our clusters experience unreasonably high memory load. To reduce the unknown-where memory cost we introduce “auto-memory-release”, thanks to @linlinhaohao888.
Currently this feature is disable by default due to severe performance downgrade during each run of memory release.
Related PR: XiaoMi/rdsn#278
Related Docs: TBD
[core] - ;tool = fastrun - cli_local = false - cli_remote = false - start_nfs = false - ;logging_factory_name = dsn::tools::hpc_logger + tcmalloc_release_rate = 1.0 # default - [tools.hpc_logger] - per_thread_buffer_bytes = 8192 [threadpool.THREAD_POOL_DEFAULT] - max_input_queue_length = 1024 [replication] + duplication_disabled = true + mem_release_interval_ms = 86400000 # 1day + mem_release_enabled = false # disable by default [pegasus.server] - updating_rocksdb_sstsize_interval_seconds = 600 + update_rdb_stat_interval = 600 [task..default] - fast_execution_in_network_thread = false