NSQ HA 1.9.1

1. update boltdb to fix view panic while close
2. optimize the delayed queue to reduce lock while doing the empty or compaction
3. optimize the memory usage.
4. use dep
5. allow avoiding fsync by default
6. speed up cluster recovery after the restart in the normal case.

NSQ HA 1.9.0

1. optimize the memory usage for nsqd ordered topic and allow change buffer size in config.
2. optimize the performance for nsqlookupd
3. fix possible stale read for etcd leadership
4. sync channel consume offset while joining ISR
5. optimize the time precision of the delayed and requeued message
6. fix INVALID command error.
7. fix possible race while migrating the topic
8. speed up consumer recovery from the error.

NSQ HA 1.8.1

1. nsqadmin support multi dc operation
2. nsqadmin support finish single message by manual
3. optimize the memory usage while empty delayed queue with large data
4. mpub now support ext info

NSQ HA 1.8.0

1. support disable write for whole nsqd node
2. optimize the lock for clean commit log data
3. channel sync optimize
4. check and create the missing partition while init topic

NSQ HA 1.6.9

1. optimize pub client stats
2. fix mode support more fix on commit log

NSQ HA 1.6.7

1. adjust req to end strategy to avoid req blocking cleaning old data
2. log adjust to reduce warning log
3. new slow consume trace log and cost time included
4. support inverse filter for consumer
5. some optimize on ephemeral channel

NSQ HA

1. ignore ephemeral channel while sync offset
2. fix leader session compare while update
3. fix delayed queue compact
4. reduce block while tag channel init or remove
5. improve migrate while removing a node
6. improve consume latency and adjust latency stats

NSQ HA

1. handle failed to notify topic meta info while init topic
2. fix build for windows
3. fix possible deadlock while deleting topic
4. fix deleted channel re-appear after restart

NSQ HA for windows test

Windows is not fully test by official, just publish a binary for easy deployment.

NSQ HA

• channel filter support based on message extend data
• adjust consume recovery speed after error/timeout
• optimize the stats info, remove clients information by default for all stats, need add needClients=true query parameter to get more stats about clients.
• fix consume offset sync for skipped ordered channel
• wakeup connection write if read timeout on client.
• optimize the lock while cleaning the topic old data
• fix the delayed queue catchup for delayed messages
• add retention_size_per_day to configure the retention size for topic