Con Kolivas
|
26c890c291
|
Send the diff from the proxy instance as soon as we get it and differentiate which proxy it comes from
|
10 years ago |
Con Kolivas
|
ab4d3a672f
|
Set notified flag to false immediately upon setting a new proxy
|
10 years ago |
Con Kolivas
|
64a3703f09
|
Move to pushing all notify data from the generator to avoid an extra message and reconnect clients once the current proxy has its first notification data
|
10 years ago |
Con Kolivas
|
836f2f3ff2
|
Avoid extra call to get subscription from generator by pushing it with the subscribe message to the stratifier
|
10 years ago |
Con Kolivas
|
db9a777767
|
Reconnect clients lazily on proxy switch once we receive another message from them and drop them if they don't disconnect on their own within a minute
|
10 years ago |
Con Kolivas
|
a80f9aa492
|
Don't miss dropping stratifier clients whose dropped flag is set when no reference count is held
|
10 years ago |
Con Kolivas
|
f0f628df84
|
Reuse known ckp variable in drop_client
|
10 years ago |
Con Kolivas
|
c5b856ac7c
|
Update notify after setting proxy before reconnecting clients
|
10 years ago |
Con Kolivas
|
32b88de36c
|
Explicitly set the proxy in the stratifier when we switch and avoid creating workbases from backup proxies
|
10 years ago |
Con Kolivas
|
14e3ed70c8
|
Don't try to act on inactive clients
|
10 years ago |
Con Kolivas
|
5a19bd0e4a
|
Send a reconnect if we detect a changed subscription on the current proxy
|
10 years ago |
Con Kolivas
|
57e7916296
|
Demote unnecessary warning in stratifier to debug
|
10 years ago |
Con Kolivas
|
d02862d1fe
|
Update subscriptions and notifications for proxies that aren't the current proxy for faster switching
|
10 years ago |
Con Kolivas
|
520b36cc70
|
Send which proxy we wish to get a subscribe or notify to/from the connector to stratifier and whether to reconnect
|
10 years ago |
Con Kolivas
|
f02a35dfcf
|
Send notready message when notify data is not ready in the generator instead of no response
|
10 years ago |
Con Kolivas
|
b7e71e1be7
|
Attach to every upstream pool in proxy mode all the time and fail over and back as needed
|
10 years ago |
Con Kolivas
|
52cd066563
|
Add proxies to the stratifier as its notified of their existence by the generator and issue reconnects without rejecting connections when a new subscribe is discovered
|
10 years ago |
Con Kolivas
|
622bd1541f
|
Make stratifier data contain a pointer to the proxy data and add a hashlist to be able to add multiple proxies, allocating only one for now
|
10 years ago |
Con Kolivas
|
9f07cc11e0
|
Export the char_entry_t type
|
10 years ago |
Con Kolivas
|
9dd3d95064
|
Fix dereference errors with dropall
|
10 years ago |
Con Kolivas
|
03b25370fe
|
Check for client ref counts in dropall
|
10 years ago |
Con Kolivas
|
745d95246c
|
Check for client ref counts in dropall
|
10 years ago |
Con Kolivas
|
9c1e5f2332
|
Send reconnect instead of dropall to stratifier when killing an upstream proxy and drop them lazily
|
10 years ago |
Con Kolivas
|
67fd11bc13
|
Convert use of %ld to PRId64 in stratifier
|
10 years ago |
Con Kolivas
|
67c00d57c6
|
Fine grain the locking for read/write sections using the upgradeable variants in the stratifier where suitable
|
10 years ago |
Con Kolivas
|
6db343fec6
|
Fix 1st 10 minute preauth to still grab suggested diff and auth the rest
|
10 years ago |
Con Kolivas
|
0a616baa46
|
Don't send any preauths for already authorised users for the first 10 minutes after their first successful auth
|
10 years ago |
Con Kolivas
|
15787d8932
|
We may as well read the poolstats when not in standalone mode as well on restart since we may expose them
|
10 years ago |
Con Kolivas
|
5805e83fe1
|
Trivial comment update
|
10 years ago |
Con Kolivas
|
0d2d81baba
|
Store and restore lastupdate stats by user and worker and use it to decay restored stats
|
10 years ago |
Con Kolivas
|
baf9a6497d
|
Store last update time by user and worker
|
10 years ago |
Con Kolivas
|
9d774a4f9b
|
Store the lastupdate time with the pool stats and decay the value if it's been longer than 2 minutes
|
10 years ago |
Con Kolivas
|
7990969af4
|
Read off pool stats on restarts in standalone mode instead of guessing the older hashrate by time bias
|
10 years ago |
Con Kolivas
|
e465fb6d6e
|
Don't try to discard passthrough clients to prevent derefs
|
10 years ago |
Con Kolivas
|
413f8f538a
|
Remove dereferencing error of accessing authorised after client may be killed
|
10 years ago |
Con Kolivas
|
57cc749ae3
|
Guarantee reliable worker count by counting all clients once their user is recognised
|
10 years ago |
Con Kolivas
|
9341ffb640
|
Highest client id may not be reliable enough to prevent duplicate ids
|
10 years ago |
Con Kolivas
|
a8d701df36
|
Discretely handle authorisation attempt to make it impossible to try and authorise twice
|
10 years ago |
Con Kolivas
|
a29c423052
|
Trivial comment change
|
10 years ago |
Con Kolivas
|
880101ccbc
|
Remove unnecessary extra list when clients are dead
|
10 years ago |
ckolivas
|
f6ac71033c
|
Disconnected and dead clients can't be holding a reference
|
10 years ago |
Con Kolivas
|
897df18c2f
|
Send a whole update to a client instead of just the diff if they've only been sending rejects
|
10 years ago |
Con Kolivas
|
b5fcff0bcb
|
Optimise method parsing for the common case of shares first and do not allow clients to try and subscribe or authorise twice
|
10 years ago |
Con Kolivas
|
d59787a0e5
|
Do lazy invalidation without parsing method
|
10 years ago |
Con Kolivas
|
12d9827ba1
|
Trivial comments
|
10 years ago |
Con Kolivas
|
2f53e6a004
|
Minor comment
|
10 years ago |
Con Kolivas
|
b1218a7cf0
|
Add config option to prioritise more clients than supporting clients with high hashrates
|
10 years ago |
Con Kolivas
|
d54a3156ad
|
Read off any cached mindiff when we do a delayed auth
|
10 years ago |
Con Kolivas
|
10613ecc76
|
Set client user_instance under lock to prevent race upon client removal
|
10 years ago |
Con Kolivas
|
d0bccb1e11
|
Remove unnecessary delay in removing dead clients with real dereference bug being found
|
10 years ago |