ckolivas
|
01c00ed3c2
|
Too much recruiting
|
10 years ago |
|
ckolivas
|
93a1414e29
|
Explicitly check for HUP in proxy_recv, avoiding an attempt at reading a line
|
10 years ago |
|
ckolivas
|
e3e83df196
|
Add a small sleep between attempts at pinging the stratifier at startup to avoid floods of messages
|
10 years ago |
|
ckolivas
|
16bd2fc352
|
Redirect from existing parent when child gets a reconnect
|
10 years ago |
|
ckolivas
|
90ab75ce24
|
Drop subproxies when the parent proxy thread exits
|
10 years ago |
|
ckolivas
|
0fa55f311f
|
Finesse what messages to send
|
10 years ago |
|
ckolivas
|
30297c3ba2
|
Revert ret < 0 check but keep recruiting of new subproxies
|
10 years ago |
|
ckolivas
|
576ac8f71a
|
Check return value of subsequent read_socket_line, disabling a subproxy if it has failed and recruit more subproxies if required
|
10 years ago |
|
ckolivas
|
bc7d59ad66
|
Recruit extra subproxies on a new parent with a small nonce size, being more verbose about proxy recruitment
|
10 years ago |
|
ckolivas
|
e6b79715cb
|
Abstract out recruit_subproxies function
|
10 years ago |
|
ckolivas
|
3fd9ead54a
|
Parse any extra data already in the cs buffer in proxy_recv before returning to the epoll loop
|
10 years ago |
|
Con Kolivas
|
8386d8aede
|
Reset the recruit count in the event of failed proxy_alive test
|
10 years ago |
|
Con Kolivas
|
d75dd5543c
|
Try async messages again
|
10 years ago |
|
Con Kolivas
|
d745e7135d
|
Remove the old proxy on a reconnect so we don't send shares nowhere
|
10 years ago |
|
Con Kolivas
|
918a6ebe0c
|
Make the stratifier tell the generator precisely how many extra connections we need
|
10 years ago |
|
Con Kolivas
|
a6ac868cd3
|
Rework subproxy recruitment to allow requests to stack but abandon them if the subproxy is not alive
|
10 years ago |
|
Con Kolivas
|
89f22deaff
|
Remove fds from the proxy recv epoll when they're being closed
|
10 years ago |
|
Con Kolivas
|
dd4b563cb9
|
Clean up and test return values of share decoding in submit_share
|
10 years ago |
|
Con Kolivas
|
484bdbaad2
|
Cosmetic
|
10 years ago |
|
Con Kolivas
|
2b6b3fdbd7
|
Look for replacement proxies in wait_best_proxy as well
|
10 years ago |
|
Con Kolivas
|
2f2d454763
|
Create new instances of proxies on reconnect instead of trying to overwrite old ones on reconnect
|
10 years ago |
|
Con Kolivas
|
14873b44b7
|
Convert proxy id to int64
|
10 years ago |
|
Con Kolivas
|
7f129e93c8
|
Revert "Revert "Rate limit proxy recruiting to one thread at a time""
This reverts commit 80c0ddb346.
Wrong revert.
|
10 years ago |
|
Con Kolivas
|
80c0ddb346
|
Revert "Rate limit proxy recruiting to one thread at a time"
This reverts commit d8eaa0b4e9.
|
10 years ago |
|
Con Kolivas
|
d8eaa0b4e9
|
Rate limit proxy recruiting to one thread at a time
|
10 years ago |
|
Con Kolivas
|
6b3cd82525
|
Revert "Use async send proc as a separate thread from workqueues"
This reverts commit 7de43b1c6e.
|
10 years ago |
|
Con Kolivas
|
51670edb61
|
Move to only reconnect the client sending shares to the dead proxy to smooth out reconnects
|
10 years ago |
|
Con Kolivas
|
7de43b1c6e
|
Use async send proc as a separate thread from workqueues
|
10 years ago |
|
Con Kolivas
|
08a997e190
|
Remove the unused msg_id field
|
10 years ago |
|
Con Kolivas
|
0dece89fa3
|
Fix const warning
|
10 years ago |
|
Con Kolivas
|
df7d17f3a9
|
Ensure we use int64 for internal jobid throughout and add the string to debug message from the upstream pool
|
10 years ago |
|
Con Kolivas
|
b3589a71f3
|
Add proxy info to verbose share result logging
|
10 years ago |
|
Con Kolivas
|
d9b27c612c
|
Add received proxy share result to verbose logging output
|
10 years ago |
|
Con Kolivas
|
af21689d67
|
Don't reconnect generator immediately after the reconnect message, waiting for the first proxy_alive test on the new instance
|
10 years ago |
|
Con Kolivas
|
7e343244ea
|
Change message slightly since we don't necessarily reconnect
|
10 years ago |
|
Con Kolivas
|
3e760f1993
|
Reconnect parent proxy even if subproxies are alive
|
10 years ago |
|
Con Kolivas
|
4d25a197e7
|
Only have one thread trying to reconnect a proxy at a time
|
10 years ago |
|
Con Kolivas
|
2cc67dab43
|
Only one workqueue will suffice in the generator when there's no conflict over resources
|
10 years ago |
|
Con Kolivas
|
6e95cd1f66
|
Revert "Use generic workqueues for the recruit subproxy function"
This reverts commit 0ecd5b5098.
|
10 years ago |
|
Con Kolivas
|
b153be29fa
|
Revert "Use generic workqueues for the proxy reconnect function"
This reverts commit e6ce49d2ef.
Holds up workqueues too long.
|
10 years ago |
|
Con Kolivas
|
41388a6e8c
|
Add proxy details to notify logging
|
10 years ago |
|
Con Kolivas
|
6b911d71d6
|
Send stratifier generic dead proxy message on submit share fail to drop all clients
|
10 years ago |
|
Con Kolivas
|
e6ce49d2ef
|
Use generic workqueues for the proxy reconnect function
|
10 years ago |
|
Con Kolivas
|
0ecd5b5098
|
Use generic workqueues for the recruit subproxy function
|
10 years ago |
|
Con Kolivas
|
23aa6a623a
|
Create two workqueues for the generator
|
10 years ago |
|
Con Kolivas
|
3f103ef67e
|
Use async send proc in the generator
|
10 years ago |
|
Con Kolivas
|
234f76f8e5
|
Force a reconnect send to clients that have been given the reconnect flag already and are requested to reconnect again
|
10 years ago |
|
Con Kolivas
|
0d39d22c16
|
Use the disable_subproxy function for all proxies including the parent and ensure we message the stratifier about all dead proxies, leaving recruitment only to on demand
|
10 years ago |
|
Con Kolivas
|
0afbf3eff2
|
Only consider a proxy dead if all subproxy connections are also dead
|
10 years ago |
|
Con Kolivas
|
396b656c82
|
Use the correct json
|
10 years ago |
