semisync replication is slow and incompatible with slave_compressed_protocol

[monty solomon]

The QPS drops from around 300 QPS to around 30 QPS when
slave_compressed_protocol is enabled running the following command on the master with two (2) slaves.

pt-heartbeat --database heartbeat --table heartbeat --update --create-table
--daemonize --interval 0.01

The state on the master is "Waiting for semi-sync ACK from slave"

master my.cnf settings

plugin-load = rpl_semi_sync_master=semisync_master.so;rpl_semi_sync_slave=semisync_slave.so
rpl_semi_sync_master_timeout = 1000000000000000000
rpl_semi_sync_master_wait_no_slave = 1
rpl_semi_sync_master_enabled = 1
rpl_semi_sync_slave_enabled = 0
slave_compressed_protocol = 1

slave my.cnf settings

plugin-load = rpl_semi_sync_master=semisync_master.so;rpl_semi_sync_slave=semisync_slave.so
rpl_semi_sync_master_timeout = 1000000000000000000
rpl_semi_sync_master_wait_no_slave = 1
rpl_semi_sync_master_enabled = 0
rpl_semi_sync_slave_enabled = 1
slave_compressed_protocol = 1

[monty solomon (monty+launchpad)]

See upstream bug https://bugs.mysql.com/bug.php?id=85382

[monty solomon (monty+launchpad)]

mysql Ver 14.14 Distrib 5.7.17-13, for Linux (x86_64) using 6.0
Server version: 5.7.17-13-log Percona Server (GPL), Release 13, Revision fd33d43
CentOS release 6.8 (Final)

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-1804