Existing user associations are not flushed

[h-homma]

Environment:

CentOS 7.6
Dovecot 2.2.36

Problem description:

When a server goes down, HOST-DOWN and HOST-FLUSH commands are issued in succession, but existing user associations are often not flushed.

When issued immediately after the first command, the second command may not be processed:

[centos@mail11 ~]$ sudo doveadm director status -a /var/run/dovecot-director/director-admin
mail server ip tag vhosts state state changed       users
192.168.250.30     100    up    -                   1
192.168.251.30     100    up    2019-08-29 19:05:14 1
[centos@mail11 ~]$ sudo nc -U /var/run/dovecot-director/director-admin -c '\
> echo -en "VERSION\tdirector-doveadm\t1\t0\n"
> read response; echo "$response" > /dev/tty
> echo -en "HOST-DOWN\t192.168.251.30\n"
> echo -en "HOST-FLUSH\t192.168.251.30\n"'
VERSION director-doveadm        1       0
[centos@mail11 ~]$ sudo doveadm director status -a /var/run/dovecot-director/director-admin
mail server ip tag vhosts state state changed       users
192.168.250.30     100    up    -                   1
192.168.251.30     100    down  2019-08-29 19:05:34 1

On the other hand, the second command seems to always be processed when issued after receiving 'OK' response for the first command:

[centos@mail11 ~]$ sudo doveadm director status -a /var/run/dovecot-director/director-admin
mail server ip tag vhosts state state changed       users
192.168.250.30     100    up    -                   1
192.168.251.30     100    up    2019-08-29 19:05:44 1
[centos@mail11 ~]$ sudo nc -U /var/run/dovecot-director/director-admin -c '\
> echo -en "VERSION\tdirector-doveadm\t1\t0\n"
> read response; echo "$response" > /dev/tty
> echo -en "HOST-DOWN\t192.168.251.30\n"
> read response; echo "$response" > /dev/tty
> echo -en "HOST-FLUSH\t192.168.251.30\n"
> read response; echo "$response" > /dev/tty'
VERSION director-doveadm        1       0
OK
OK
[centos@mail11 ~]$ sudo doveadm director status -a /var/run/dovecot-director/director-admin
mail server ip tag vhosts state state changed       users
192.168.250.30     100    up    -                   2
192.168.251.30     100    down  2019-08-29 19:13:08 0

I think command responses should always be received and checked for reliable operation.