pongbot does not reconnect #107
Comments
|
If you can reproduce then we probably do have a bug. Could use a test/fix. Thanks. |
|
Thank you. I can reproduce that WiFi-case error on macOS and Linux (Raspberry PI). I don't know which layer should raise the error. In this case, the client looks still connected, So, I think pong-bot needs to detect the disconnection using a kind of watchdog-mechanism |
|
Try with a bot that's implemented on top of https://github.com/dblock/slack-ruby-bot-server, I am pretty sure it has reconnect logic that handles it. For example https://github.com/dblock/slack-market. |
|
Thanks for your kind suggestion. I'll try it. Let me confirm one more thing. |
|
In theory we should have complete reconnect logic here. So what you described initially still looks like a bug to me, but you'll have to dig through it and possibly PR a fix. |
|
Maybe, this issue is related with faye/websocket-driver-ruby#42 and celluloid/celluloid-io#176. Also, slack-ruby/slack-ruby-client#96 looks very similar. @tikh reproduced the same situation by turning off his WiFi. One thing different was periodic pining created a catchable exception, while my pongbot is too shy to sending pings. I'm not sure I should make pongbot chatty, or hack celluloid-io. |
|
I settled to add such ping thread to my pongbot: class PingThread < SlackRubyBot::Server
on 'hello' do |client, data|
puts "PingThread: Hello!"
@reconnect_count ||= 0
if @ping_thread
@ping_thread.exit
@reconnect_count += 1
end
@ping_thread = Thread.new do
sequence = 1
loop do
client.ping({id: sequence})
sleep 30
sequence += 1
end
end
end
endThis thread emits |
|
If you're using Celluloid you might want to change this not to use a thread. You can add |
|
I'd love for this to be turned into a feature, something that checks bot connections periodically and is built in. |
|
Thanks for the comment. I've tried to replace thread by class PingThread < SlackRubyBot::Server
extend Celluloid
on 'hello' do |client, data|
puts "PingThread: Hello!"
@reconnect_count ||= 0
if @timer
@timer.cancel
@reconnect_count += 1
end
sequence = 1
@timer = every 30 do
client.ping({id: sequence})
puts "PingThread: ping #{sequence}, reconnect: #{@reconnect_count}."
sequence += 1
end
end
endI had to use |
|
Looks good to me. |
|
Thanks. I hope this mechanism becomes a built-in feature. |
|
I have similar problem, but it's worse as I understand. I have bot running without bot server, just on celluloid by but no Now I fixed this by listening for the def run
instance.on 'hello', ->(client, data) do
logger.debug "PingThread: Hello!"
@reconnect_count ||= 0
if @timer
@timer.cancel
@reconnect_count += 1
end
@sequence = 1
@timer = every 15 do
if @pong_recieved == false
@pong_recieved = nil
logger.warn 'Pong for previous ping not recieved, restarting.'
instance.instance_variable_set('@client', nil)
instance.restart!
next
end
client.ping(id: @sequence)
logger.debug "PingThread: ping #{@sequence}, reconnect: #{@reconnect_count}."
@pong_recieved = false
end
end
instance.on 'pong', ->(client, data) do
logger.debug "PingThread: Pong received!"
if data['reply_to'] == @sequence
@pong_recieved = true
@sequence += 1
end
end
super
end |
|
I've recently been doing this. I definitely think we're not properly handling a server-side disconnect, see slack-ruby/slack-ruby-client#208, with someone at Slack looking at helping us. |
|
To tell the truth, my first version was very similar to @gsmetal 's one like below. class PingThread < SlackRubyBot::Server
extend Celluloid
PONG_ID_QUEUE = []
def self.pong_received?(id)
!PONG_ID_QUEUE.delete(id).nil?
end
on 'hello' do |client, data|
puts "Ping Thread HELLO!"
PONG_ID_QUEUE.clear
@reconnect_count ||= 0
if @timer
@timer.cancel
@reconnect_count += 1
end
sequence = 1
@timer = every 30 do
client.ping({id: sequence})
puts "ping #{sequence}, reconnect: #{@reconnect_count}."
sleep 3
raise Errno::ETIMEDOUT unless pong_received?(sequence)
sequence += 1
end
end
on 'pong' do |client, data|
sequence = data[:reply_to].to_i
puts "pong #{sequence}"
PONG_ID_QUEUE << sequence
end
end |
|
Hi, I am trying to use the simple pong example but I keep getting this error: |
|
Resolved via slack-ruby/slack-ruby-client#208 |
Hi, I'm playing with your slack-ruby-bot in my holiday.
I started from pongbot in README and talked with the bot, it's 👍
However, after several minutes of network disconnection,
I noticed that pongbot does not try to reconnect to Slack.
For example, turn-off my WiFi network, wait several minutes, and then reconnect WiFi,
pongbot is left disconnected even if WiFi is back.
Terminal does not print any line during that experiment.
I believe
handle_exceptionsin server.rb kicksrestart!.However, it seems not to catch any exception in this situation.
Any hint to add reconnect feature to pongbot ?
I'm using Ruby 2.3.3 and macOS Sierra and slack-ruby-bot 0.9.
Here is the Gemfile.lock for the pongbot
The text was updated successfully, but these errors were encountered: