multi: Add Balancer for round-robin redundancy #1078

maicalal · 2018-02-24T17:28:19Z

Added a RoundRobinBalancer to do a round robin over a list of
hosts using a picker. List of hosts to be set in connection config
thats used to create a new Client.
Changes in file roundrobin.go:

Picker
- responsible to pick the next usable connection
RoundrobinBalancer
- holds a picker instance to get the next connection
- maintains a map for storing the current connection
  state for a host
- maintains a map of socket connections by host

Changes in file infrastructure.go:

Client now has a reference to roundrobinbalancer
Changes to following workflows
- Creating New client
  - Does a rrrb.Get() to have one ready connection
- wsInHandler
  - iterates over all active connections to check for
    messages
- wsOutHandler
  - picks a connection using balancer to send out
    message
- wsReconnectHandler
  - gets the disconnected connections from balancer
    and attempts to reconnect.
    If all were disconnected, it would flag to call start()
    This fixes rpcclient: Add round-robin redundancy. #884

davecgh

Thanks for the PR! This should be quite nice for some services that use rpcclient to connect to multiple backends.

This is just a first pass where I focused on style-related issues such as consistency and things that don't follow the Effective Go Guidelines.

I'll spend more time digging into the logic and testing it in subsequent reviews, but I wanted to go ahead and get most of these more minor things out of the way first.

rpcclient/infrastructure.go

rpcclient/roundrobin.go

davecgh · 2018-03-05T21:13:05Z

Just an FYI that I will be doing some thorough testing and review on this after the upcoming release. While we always try to make sure everything is well tested and functioning before it goes to master, this is doubly the case when a release is looming, especially for things that have huge implications such as the RPC client upon which wallets, stakepools, and mining software rely.

maicalal · 2018-03-06T14:08:35Z

Thanks @davecgh , will wait for the feedback.

maicalal · 2018-04-28T02:28:28Z

hi @davecgh any luck :) ...

rpcclient/infrastructure.go

rpcclient/roundrobin.go

rpcclient/infrastructure.go

rpcclient/roundrobin.go

maicalal · 2018-05-08T20:23:08Z

Thanks @davecgh , working on it...

davecgh · 2018-05-09T19:50:45Z

Great. Thanks! Please either leave a comment on the PR or ping me via one of the chat apps once you've updated as github isn't the best about notifying on updates to PRs otherwise.

maicalal · 2018-05-13T20:04:15Z

Hi @davecgh ,
Comments got screwed up in my last checkin. Here is the same:

Incorporating davecgh’s feedback

WsInHandler to be per wsConn
Moved the Check and setting flag to indicate need for client restart to balancer.close()
Added balancer.CloseAll()
Other formatting changes

rpcclient/roundrobin.go

rpcclient/infrastructure.go

rpcclient/roundrobin.go

rpcclient/infrastructure.go

Please enter the commit message for your changes. Lines starting

Updated comments and changed method names as per guidelines. Made one bugfix in NextConn() function where it was doing multiple lookups.

…ed the Check and setting flag to indicate need for client restart to balancer.close() - Added balancer.CloseAll() - Other formatting changes

- Balancer remembers conn used for a notification once used. Hence, no round-robin for these notification methods. - Moved Balancer interface to balancer.go - Fixed other comments from davecgh

Ran rpctest locally and goes fine

jrick · 2018-06-15T12:58:58Z

This change strikes me as being backwards. I would expect a load balancer to be created from many clients, and not a single client load balancing over multiple servers. It should be possible to create a load balancing client in another package using only the exported features of the rpcclient package. Creating a new type to handle load balanced requests also means we can choose to not introduce any of the RPC methods that keep state and can't be load balanced. Is there some reason I'm missing why it wasn't designed this way?

maicalal · 2018-06-21T10:16:09Z

Hi @jrick this was the base requirement for issue #884 .
It made sense to have a pure client side balancer where it has multiple servers that client can connect to, so went ahead with the changes.
Moreover, a load balancer over multiple clients, would change its meaning and would be more like a message broker in a pub/sub paradigm.

maicalal · 2018-10-10T09:18:05Z

Hi @dnldd , have resolved all previous comments by @davecgh .
Also have fixed the issues you faced.
The flow is as below:
Say con1 and con2 are specified to be used by balancer.
Initial connection (i.e. first successful connection):

code would make 50 attempts for each connection, then mark it as shutdown.

Reconnection:

after a successful initial connection, if either one of con1 and con2, disconnects,
this would result into making reconnection attempts as per the settings.
during this phase, if there are no active connections, code will not execute any sendcmd (this is same as earlier, only difference is, earlier it would have used state of "disconnect" channel, but now we would need to rely on available active connections)

dnldd · 2018-10-10T21:43:36Z

noted, I plan on reviewing soon. Will keep you updated.

dnldd · 2018-10-23T21:07:48Z

here is the config I tested with:

	connCfg := &rpcclient.ConnConfig{
		User:                 "user",
		Pass:                 "pass,
		DisableAutoReconnect: false,
		Certificates:         certs,
		HostAddresses: []rpcclient.HostAddress{
			rpcclient.HostAddress{
				Host:     "127.0.0.1:19110",
				Endpoint: "ws",
			},
			rpcclient.HostAddress{
				Host:     "127.0.0.1:19111",
				Endpoint: "ws",
			},
			rpcclient.HostAddress{
				Host:     "127.0.0.1:19109",
				Endpoint: "ws",
			},
		},

I initially started with :19109(live), :19110(shutdown) and 19111(shutdown). I then killed :19109 and started :19110, resulting in :19109(shutdown), :19110(live) and 19111(shutdown). Here is the output for that:

2018-10-23 20:53:16.248 [INF] RPCC: Connection state update for host 127.0.0.1:19110: CONNECTING
2018-10-23 20:53:16.250 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:16.250 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:1
2018-10-23 20:53:16.250 [INF] RPCC: Connection state update for host 127.0.0.1:19111: CONNECTING
2018-10-23 20:53:16.250 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:16.250 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:1
2018-10-23 20:53:16.250 [INF] RPCC: Connection state update for host 127.0.0.1:19109: CONNECTING
2018-10-23 20:53:16.283 [INF] RPCC: Connection state update for host 127.0.0.1:19109: READY
2018-10-23 20:53:16.283 [INF] RPCC: Balancer: Established connection to RPC server 127.0.0.1:19109
2018-10-23 20:53:16.283 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:16.284 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:16.284 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:2
2018-10-23 20:53:16.284 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:16.284 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:2
2018-10-23 20:53:16.284 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:16.284 [INF] RPCC: Balancer: Will be using 127.0.0.1:19109 for notification: notifyblocks
2018-10-23 20:53:16.287 [INF] RPCC: NotifyBlocks: Registration Complete
2018-10-23 20:53:21.293 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:21.293 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:3
2018-10-23 20:53:21.293 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:21.293 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:3
2018-10-23 20:53:21.293 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:21.294 [INF] RPCC: Block count: 55353
2018-10-23 20:53:26.289 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:26.289 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:4
2018-10-23 20:53:26.289 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:26.289 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:4
2018-10-23 20:53:26.289 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:26.290 [INF] RPCC: Block count: 55353
2018-10-23 20:53:31.292 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:31.292 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:5
2018-10-23 20:53:31.292 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:31.292 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:5
2018-10-23 20:53:31.293 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:31.293 [INF] RPCC: Block count: 55353
2018-10-23 20:53:36.290 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:36.290 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:6
2018-10-23 20:53:36.291 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:36.291 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:6
2018-10-23 20:53:36.291 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:36.291 [INF] RPCC: Block count: 55353
2018-10-23 20:53:41.294 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:41.294 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:7
2018-10-23 20:53:41.294 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:41.294 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:7
2018-10-23 20:53:41.295 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:41.295 [INF] RPCC: Block count: 55353
2018-10-23 20:53:46.289 [INF] RPCC: Connection state update for host 127.0.0.1:19110: IDLE
2018-10-23 20:53:46.289 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused, attempt:8
2018-10-23 20:53:46.289 [INF] RPCC: Connection state update for host 127.0.0.1:19111: IDLE
2018-10-23 20:53:46.289 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused, attempt:8
2018-10-23 20:53:46.289 [INF] RPCC: Balancer: Connection pick for RPC 127.0.0.1:19109
2018-10-23 20:53:46.290 [INF] RPCC: Block count: 55353
2018-10-23 20:53:50.696 [ERR] RPCC: Websocket receive error from 127.0.0.1:19109: websocket: close 1006 (abnormal closure): unexpected EOF
2018-10-23 20:53:50.697 [INF] RPCC: Connection state update for host 127.0.0.1:19109: DISCONNECTED
2018-10-23 20:53:50.698 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:53:50.698 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 10s
2018-10-23 20:54:00.701 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:54:00.701 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 15s
2018-10-23 20:54:15.702 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:54:15.702 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 20s
2018-10-23 20:54:35.708 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:54:35.708 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 25s
2018-10-23 20:55:00.713 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:55:00.713 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 30s
2018-10-23 20:55:30.719 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:55:30.719 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 35s
2018-10-23 20:56:05.725 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:56:05.725 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 40s
2018-10-23 20:56:45.728 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:56:45.728 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 45s
2018-10-23 20:57:30.734 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-10-23 20:57:30.734 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 50s
...

From the logs, it's continuously retrying :19109 when at this point it should be trying to connect to :19109, :19110 and :19111 like it did on startup. On a disconnect cycle though all host addresses for a connection, if none is found you back off the retry cycle eg. Retrying all host addresses in 10s, 20s, 30s, 40s ...

I'm yet to peruse your updates on the comments, expect a review on that soon.

maicalal · 2018-10-24T14:17:21Z

Hi @dnldd ,
The reconnect handler kicks in once a connection gets disconnected. As we can see in the logs
Retrying connection to 127.0.0.1:19109 in ...s
The connections for :19110 and :19111 never got connected, so they do not get into the reconnect workflow.

Do you want me to change the reconnect handler to work on these as well ?
This will be a small change. Didn't do it earlier because I thought they do not qualify, as they never got connected.

dnldd · 2018-10-26T20:42:25Z

Yeah, let's do that. On a disconnection retry all host addresses.

connections in the HostAddres list that are not in connected state.

maicalal · 2018-11-06T04:00:45Z

hi @dnldd , have updated the code accordingly.
Please try with latest.

dnldd

The issues highlighted here are not exhaustive, use them to correct similar comment related issues throughout the pr. I noticed variable names were still being used in comments. I also do not think its necessary to refer to a connection as ws connection particularly because of the abbreviation ws, just connection should be fine.

Will be testing the updates made shortly.

rpcclient/balancer.go

dnldd · 2018-11-13T02:40:02Z

When all hosts are unavailable, the initial retry cycle does not back off.

2018-11-13 02:21:26.137 [INF] RPCC: Connection state update for host 127.0.0.1:9110: IDLE
2018-11-13 02:21:26.137 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9110: dial tcp 127.0.0.1:9110: connect: connection refused, attempt:1
2018-11-13 02:21:26.137 [INF] RPCC: Connection state update for host 127.0.0.1:9111: CONNECTING
2018-11-13 02:21:26.138 [INF] RPCC: Connection state update for host 127.0.0.1:9111: IDLE
2018-11-13 02:21:26.138 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9111: dial tcp 127.0.0.1:9111: connect: connection refused, attempt:1
2018-11-13 02:21:26.138 [INF] RPCC: Connection state update for host 127.0.0.1:9109: CONNECTING
2018-11-13 02:21:26.139 [INF] RPCC: Connection state update for host 127.0.0.1:9109: IDLE
2018-11-13 02:21:26.139 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9109: dial tcp 127.0.0.1:9109: connect: connection refused, attempt:1
2018-11-13 02:21:26.139 [INF] RPCC: Connection state update for host 127.0.0.1:9110: IDLE
2018-11-13 02:21:26.149 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9110: dial tcp 127.0.0.1:9110: connect: connection refused, attempt:2
2018-11-13 02:21:26.151 [INF] RPCC: Connection state update for host 127.0.0.1:9111: IDLE
2018-11-13 02:21:26.151 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9111: dial tcp 127.0.0.1:9111: connect: connection refused, attempt:2
...
2018-11-13 02:21:27.052 [INF] RPCC: Connection state update for host 127.0.0.1:9110: IDLE
2018-11-13 02:21:27.052 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9110: dial tcp 127.0.0.1:9110: connect: connection refused, attempt:47
2018-11-13 02:21:27.054 [INF] RPCC: Connection state update for host 127.0.0.1:9111: IDLE
2018-11-13 02:21:27.089 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9111: dial tcp 127.0.0.1:9111: connect: connection refused, attempt:47
2018-11-13 02:21:27.090 [INF] RPCC: Connection state update for host 127.0.0.1:9109: IDLE
2018-11-13 02:21:27.090 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9109: dial tcp 127.0.0.1:9109: connect: connection refused, attempt:47
2018-11-13 02:21:27.091 [INF] RPCC: Connection state update for host 127.0.0.1:9110: IDLE
2018-11-13 02:21:27.091 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9110: dial tcp 127.0.0.1:9110: connect: connection refused, attempt:48
2018-11-13 02:21:27.091 [INF] RPCC: Connection state update for host 127.0.0.1:9111: IDLE
2018-11-13 02:21:27.091 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9111: dial tcp 127.0.0.1:9111: connect: connection refused, attempt:48
2018-11-13 02:21:27.092 [INF] RPCC: Connection state update for host 127.0.0.1:9109: IDLE
2018-11-13 02:21:27.113 [INF] RPCC: Balancer: Failed to make initial connection to 127.0.0.1:9109: dial tcp 127.0.0.1:9109: connect: connection refused, attempt:48
...

50 retries for each host are exhausted in a matter of seconds as a result. Is there any reason why you're not backing off here as compared to a dropped connection? Here is an example:

...
2018-11-13 02:26:24.092 [INF] RPCC: Retrying connection to 127.0.0.1:19110 in 10s
2018-11-13 02:26:34.095 [INF] RPCC: Failed to connect to 127.0.0.1:19111: dial tcp 127.0.0.1:19111: connect: connection refused
2018-11-13 02:26:34.095 [INF] RPCC: Retrying connection to 127.0.0.1:19111 in 10s
2018-11-13 02:26:44.101 [INF] RPCC: Failed to connect to 127.0.0.1:19109: dial tcp 127.0.0.1:19109: connect: connection refused
2018-11-13 02:26:44.101 [INF] RPCC: Retrying connection to 127.0.0.1:19109 in 10s
2018-11-13 02:26:54.104 [INF] RPCC: Failed to connect to 127.0.0.1:19110: dial tcp 127.0.0.1:19110: connect: connection refused
2018-11-13 02:26:54.104 [INF] RPCC: Retrying connection to 127.0.0.1:19110 in 15s
...

maicalal · 2018-11-15T10:55:47Z

Hi @dnldd , I assumed those which never got connected, need to be handled differently.
Looks like am wrong here.
Will fix the code to match the backoff strategy as in the case of dropped connections.
Working on it.

maicalal · 2018-12-08T02:22:10Z

Apologies for the delay. Its due to office load and will get back to this by 3rd week of dec.

Retry on initial connection to follow backoff as during reconnect. Fixed few bugs for getting next connection during reconnects in progress.

during initial connection.

causing tests to timeout as code tries to reconnect.

maicalal · 2018-12-23T20:33:02Z

Hi @dnldd have incorporated your suggestions. Please try with latest.

maicalal · 2019-01-18T17:19:52Z

Hi @dnldd, please let me know if you want me to make any more changes...

rpcclient/roundrobin.go

dajohi

have you built and ran this with go build -race?

jrick · 2019-01-18T18:59:16Z

FYI - I have no plans on pulling this change into dcrwallet due to my aforementioned concerns of its design. Having multiple RPC servers to sync the wallet with is useful, but the syncer should have direct knowledge of each of these servers (similar to how the SPV syncer directly handles all connected peers).

I was already planning on removing the rpcclient dependency from dcrwallet entirely for other reasons (it's already more complex to work with than it should be), but this would just accelerate that process.

maicalal · 2019-02-01T02:33:51Z

have you built and ran this with go build -race?

Hi @dajohi, had run with -race but it did not detect the code you pointed out in picker.
Maybe because it goes into this if block only once, then it errors out saying no connection available.
Will checkin the changes in sometime.

davecgh · 2019-10-26T01:15:48Z

This never really got any traction, since, as it turns out, all primary consumers wanted to control the connection management themselves. One of them in particular, the wallet, decided to move away from rpcclient altogether. Consequently, while I really am not happy about closing this without merging after all this effort and time, none of the consumers want it, and it would just end up being code that has to be maintained for no benefit. Closing.

davecgh requested changes Feb 24, 2018

View reviewed changes

davecgh mentioned this pull request Mar 5, 2018

rpcclient: Add round-robin redundancy. #884

Closed

davecgh added this to the 1.3.0 milestone Apr 30, 2018

davecgh requested changes May 3, 2018

View reviewed changes

davecgh requested changes May 26, 2018

View reviewed changes

maicalal force-pushed the master branch 4 times, most recently from ce6983d to 2e7f3a7 Compare June 7, 2018 19:12

maicalal added 8 commits June 8, 2018 08:41

Balancer changes without test case

c9f10c0

fixing comments

bfdfb6d

Fixes for Disconnect, Reconnect and Shutdown workflow

a080f25

Please enter the commit message for your changes. Lines starting

Fixing build warnings

3f53dc7

Ran gofmt -s locally. Was relying on visual code.

054d20b

Fixing review comments from davecgh.

c35f77a

Updated comments and changed method names as per guidelines. Made one bugfix in NextConn() function where it was doing multiple lookups.

Incorporating davecgh’s feedback - WsInHandler to be per wsConn - Mov…

61b31d4

…ed the Check and setting flag to indicate need for client restart to balancer.close() - Added balancer.CloseAll() - Other formatting changes

Implemented changes as suggested by davecgh and jrick.

0e73daf

- Balancer remembers conn used for a notification once used. Hence, no round-robin for these notification methods. - Moved Balancer interface to balancer.go - Fixed other comments from davecgh

maicalal force-pushed the master branch from 2e7f3a7 to 0e73daf Compare June 8, 2018 03:14

Changing import path to use "github.com/gorilla/websocket"

b72f2a8

Ran rpctest locally and goes fine

davecgh mentioned this pull request Jun 14, 2018

rpcclient: Implement fmt.Stringer for Client #1298

Merged

davecgh removed this from the 1.3.0 milestone Aug 1, 2018

Incorporating dnldd's feedback, i.e. upon disconnect, retry all

4e09409

connections in the HostAddres list that are not in connected state.

dnldd suggested changes Nov 13, 2018

View reviewed changes

davecgh modified the milestones: 1.4.0, 1.5.0 Nov 14, 2018

maicalal added 3 commits December 23, 2018 13:18

Updated comments as per guidelines.

6c71e08

Retry on initial connection to follow backoff as during reconnect. Fixed few bugs for getting next connection during reconnects in progress.

Fixing testcase: should respect DisableAutoReconnect

e2693ee

during initial connection.

TestClientStringer test was not setting DisableAutoReconnect

e5502ab

causing tests to timeout as code tries to reconnect.

dajohi reviewed Jan 18, 2019

View reviewed changes

rpcclient/roundrobin.go Show resolved Hide resolved

dajohi reviewed Jan 18, 2019

View reviewed changes

rpcclient/roundrobin.go Outdated Show resolved Hide resolved

dajohi requested changes Jan 18, 2019

View reviewed changes

maicalal added 3 commits February 1, 2019 09:22

Incorporating dajohi's review comments.

d5969d5

Missed one of the review comments.

9563961

Deleting run_tests.sh as it got accidentally added

92a580a

davecgh force-pushed the master branch from 2713aab to 4e7f080 Compare March 19, 2019 15:27

davecgh removed this from the 1.5.0 milestone Oct 14, 2019

davecgh closed this Oct 26, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

multi: Add Balancer for round-robin redundancy #1078

multi: Add Balancer for round-robin redundancy #1078

maicalal commented Feb 24, 2018

davecgh left a comment

davecgh commented Mar 5, 2018 •

edited

Loading

maicalal commented Mar 6, 2018

maicalal commented Apr 28, 2018

maicalal commented May 8, 2018

davecgh commented May 9, 2018 •

edited

Loading

maicalal commented May 13, 2018 •

edited

Loading

jrick commented Jun 15, 2018 •

edited

Loading

maicalal commented Jun 21, 2018

maicalal commented Oct 10, 2018 •

edited

Loading

dnldd commented Oct 10, 2018

dnldd commented Oct 23, 2018 •

edited

Loading

maicalal commented Oct 24, 2018

dnldd commented Oct 26, 2018

maicalal commented Nov 6, 2018

dnldd left a comment

dnldd commented Nov 13, 2018

maicalal commented Nov 15, 2018

maicalal commented Dec 8, 2018

maicalal commented Dec 23, 2018

maicalal commented Jan 18, 2019

dajohi left a comment

jrick commented Jan 18, 2019 •

edited

Loading

maicalal commented Feb 1, 2019

davecgh commented Oct 26, 2019

multi: Add Balancer for round-robin redundancy #1078

multi: Add Balancer for round-robin redundancy #1078

Conversation

maicalal commented Feb 24, 2018

davecgh left a comment

Choose a reason for hiding this comment

davecgh commented Mar 5, 2018 • edited Loading

maicalal commented Mar 6, 2018

maicalal commented Apr 28, 2018

maicalal commented May 8, 2018

davecgh commented May 9, 2018 • edited Loading

maicalal commented May 13, 2018 • edited Loading

jrick commented Jun 15, 2018 • edited Loading

maicalal commented Jun 21, 2018

maicalal commented Oct 10, 2018 • edited Loading

dnldd commented Oct 10, 2018

dnldd commented Oct 23, 2018 • edited Loading

maicalal commented Oct 24, 2018

dnldd commented Oct 26, 2018

maicalal commented Nov 6, 2018

dnldd left a comment

Choose a reason for hiding this comment

dnldd commented Nov 13, 2018

maicalal commented Nov 15, 2018

maicalal commented Dec 8, 2018

maicalal commented Dec 23, 2018

maicalal commented Jan 18, 2019

dajohi left a comment

Choose a reason for hiding this comment

jrick commented Jan 18, 2019 • edited Loading

maicalal commented Feb 1, 2019

davecgh commented Oct 26, 2019

davecgh commented Mar 5, 2018 •

edited

Loading

davecgh commented May 9, 2018 •

edited

Loading

maicalal commented May 13, 2018 •

edited

Loading

jrick commented Jun 15, 2018 •

edited

Loading

maicalal commented Oct 10, 2018 •

edited

Loading

dnldd commented Oct 23, 2018 •

edited

Loading

jrick commented Jan 18, 2019 •

edited

Loading