I have a very basic application for flasks in uwsgi, controlled by signals under a supervisor. I use cheaper to scale workers and faced with a very worrying situation. uwsgi kills the oldest worker even when it processes the request by calling 500.
How to prevent uwsgi from killing workers processing the request for the sake of saving?
Any help / tips are deeply appreciated. Found another similar post without an answer: UWSGI killed workers too fast
Configuration: uwsgi version: 2.0.4
workers = 30
cheaper = 2
cheaper-step = 3
cheaper-algo = backlog
cheaper-overload = 2
Logs:
[pid: 15601|app: 0|req: 13/78] 10.83.9.61 () {28 vars in 368 bytes} [Fri Sep 30 02:00:34 2016] POST /xxxxxxxxxxxx => generated 257 bytes in 151 msecs (HTTP/1.1 200) 2 headers in 72 bytes (1 switches on core 0)
[pid: 15601|app: 0|req: 14/79] 10.83.9.61 () {28 vars in 368 bytes} [Fri Sep 30 02:00:35 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 649 msecs (HTTP/1.1 200) 2 headers in 72 bytes (1 switches on core 0)
[pid: 15600|app: 0|req: 21/80] 10.83.9.62 () {28 vars in 369 bytes} [Fri Sep 30 03:30:08 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 2696 msecs (HTTP/1.1 200) 2 headers in 72 bytes (3 switches on core 0)
Respawned uWSGI worker 1 (new pid: 21009)
Respawned uWSGI worker 4 (new pid: 21010)
Respawned uWSGI worker 5 (new pid: 21011)
[pid: 15601|app: 0|req: 15/81] 10.83.9.62 () {28 vars in 370 bytes} [Fri Sep 30 03:30:09 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 3046 msecs (HTTP/1.1 200) 2 headers in 72 bytes (2 switches on core 0)
[pid: 15600|app: 0|req: 22/82] 10.83.9.63 () {28 vars in 370 bytes} [Fri Sep 30 03:30:10 2016] POST /xxxxxxxxxxxx => generated 0 bytes in 3347 msecs (HTTP/1.1 500) 2 headers in 127 bytes (7 switches on core 0)
worker 2 killed successfully (pid: 15600) <------------- pid 15600 was interrupted, returned 500 (line above)
uWSGI worker 2 cheaped.
[pid: 15601|app: 0|req: 16/83] 10.83.9.62 () {28 vars in 369 bytes} [Fri Sep 30 03:30:12 2016] POST /xxxxxxxxxxxx => generated 0 bytes in 2560 msecs (HTTP/1.1 500) 2 headers in 127 bytes (3 switches on core 0)
[pid: 21010|app: 0|req: 19/84] 10.83.9.63 () {28 vars in 369 bytes} [Fri Sep 30 03:30:12 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 2931 msecs (HTTP/1.1 200) 2 headers in 72 bytes (4 switches on core 0)
worker 3 killed successfully (pid: 15601) <------------- pid 15601 was interrupted, returned 500 (line above)
uWSGI worker 3 cheaped.
[pid: 21011|app: 0|req: 22/85] 10.83.9.61 () {28 vars in 370 bytes} [Fri Sep 30 03:30:12 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 3236 msecs (HTTP/1.1 200) 2 headers in 72 bytes (4 switches on core 0)
[pid: 21009|app: 0|req: 7/86] 10.83.9.61 () {28 vars in 370 bytes} [Fri Sep 30 03:30:12 2016] POST /xxxxxxxxxxxx => generated 0 bytes in 3532 msecs (HTTP/1.1 500) 2 headers in 127 bytes (4 switches on core 0)
worker 1 killed successfully (pid: 21009) <------------- pid 15601 was interrupted, returned 500 (line above)
uWSGI worker 1 cheaped.
[pid: 21011|app: 0|req: 23/87] 10.83.9.63 () {28 vars in 369 bytes} [Fri Sep 30 03:30:15 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 1614 msecs (HTTP/1.1 200) 2 headers in 72 bytes (1 switches on core 0)
Respawned uWSGI worker 1 (new pid: 21013)
Respawned uWSGI worker 2 (new pid: 21014)
Respawned uWSGI worker 3 (new pid: 21015)
[pid: 21011|app: 0|req: 24/88] 10.83.9.61 () {28 vars in 369 bytes} [Fri Sep 30 03:30:17 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 2546 msecs (HTTP/1.1 200) 2 headers in 72 bytes (3 switches on core 0)
[pid: 21010|app: 0|req: 20/89] 10.83.9.62 () {28 vars in 370 bytes} [Fri Sep 30 03:30:15 2016] POST /xxxxxxxxxxxx => generated 0 bytes in 4845 msecs (HTTP/1.1 500) 2 headers in 127 bytes (6 switches on core 0)
worker 4 killed successfully (pid: 21010) <------------- pid 15601 was interrupted, returned 500 (line above)
uWSGI worker 4 cheaped.
[pid: 21015|app: 0|req: 17/90] 10.83.9.61 () {28 vars in 369 bytes} [Fri Sep 30 03:30:18 2016] POST /xxxxxxxxxxxx => generated 186 bytes in 2402 msecs (HTTP/1.1 200) 2 headers in 72 bytes (3 switches on core 0)
worker 5 killed successfully (pid: 21011) <------------- pid 15601 was interrupted, returned 500 (line above)
source
share