thank you very much!
is it possible to put --log-drain several times
with different REGEX's?
also when 1.3 release expected?
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
I'am interested too, how good is built in HTTP server in 1.2.6
compared to nginx? as on some nodes, we don't need NGINX at all and
using it only to proxy requests to uwsgi.
___
uWSGI mailing list
uWSGI@lists.unbit.it
Hello uWSGI,
I'am trying to read file in async mode, and getting some errors. do
I do something wrong?
fp = open('somefile', 'a')
uwsgi.wait_fd_write(fp.fileno(), 1)
uwsgi.suspend()
fp.write(msg+'\n')
uwsgi.wait_fd_write(fp.fileno(), 1)
uwsgi.suspend()
fp.flush()
# some
Hello Roberto,
but how to close it correctly?
I have 2 nodes. one just writes into own NODE, another reads from this
node. READER node generates this errors, if there are no values or
value was READ second time.
so how I should read them carefully and close instance? all I have it's
sorry, I got that it's just feature ))
but it will be great if you'll suppress these error messages, just not
to fill all logs with them. at least in cache_exists.
also what u think about idea, to add to the LOG's param, some array or
list with keywords, to ignore some lines in LOG's?
so in
Hello Roberto,
maybe you should do something like cache_close() to correctly close
it? as in other way it's hard to use if time between cache write/hit
is minimal. It's just breaks values in cache, so reader can't read
them later. or I'am doing something wrong?
Hello Roberto,
do u know if it's possible to work in uwsgi async mode with pylibmc library
for memcache?
I see there in documentation
no_block
Enables asychronous I/O. This is the fastest transport available for
storage functions.
http://sendapatch.se/projects/pylibmc/behaviors.html
Hello Uwsgi,
I'am getting this error
recvmsg(): Operation now in progress [proto/uwsgi.c line 40]
Sun Aug 26 03:50:27 2012 - error parsing request
when I'am trying to read with cache_get or cache_exists the key what
doesn't exist. is it normal? how to suppress it?
uwsgi --version
1.2.5
, that master process is separate from main
application? so I can put different code for master process? if yes,
any example how to do it?
--
Best regards,
Test157mailto:test...@smail.ee
___
uWSGI mailing list
uWSGI
I want to create global variable (list or dic object) at
init_proc, when uwsgi starts it
will be populated from database with data, and all processes will use it only
as
READ ONLY object instead of hiting DB, but I want listen postgresql for
NOTIFY event, and
repopulate this
Hello Roberto,
I'am getting this error while using it on the same machine, so I don't
use it over the network.
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hi everyone.
I have 2 uwsgi instances
1. the first one is running at 127.0.0.1:10001 with cache = 1000
(reads/writes
cache)
2.the secondisrunning at 127.0.0.1:10002 with cache = 1
(reads/deletes values only from the first instance)
when I'am trying to access cache,
chdir() to /home/some/path
closing all non-uwsgi socket fds 2 (_SC_OPEN_MAX = 32000)...
found fd 3 mapped to socket 0 (127.0.0.1:10008)
running /home/virt/some/path/uwsgi
execvp(): Bad address [master.c line 784]
getting this error, when reload file touched
running django
don't have
with this problem.
Search for solutions in Django bugs list.
Look like pgbouncer also may help
--
Best regards,
Test157mailto:test...@smail.ee
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin
hi,
what is the best practice for today to use mysql in async mode with
uwsgi? I know mysql driver doesn't support it, but what workaround
you would recommend?
___
uWSGI mailing list
uWSGI@lists.unbit.it
Hello Evan,
as far as I know pypy not yet compatible with uwsgi.
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Roberto,
i'am getting lot of async queue is full in logs
could you explain what does it mean? as I see there are no code 500
pages, but are so records in the logs
___
uWSGI mailing list
uWSGI@lists.unbit.it
.
--
Roberto De Ioris
http://unbit.it
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
--
Best regards,
Test157mailto:test...@smail.ee
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
--
Best regards,
Test157mailto:test...@smail.ee
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman
Hello,
how to echo to console with
1. logs enabled
2. logs disabled
Raise excpetion is not very comfy to use )) usual print doesn't work
from the modules, and logging doesn't work for me at all.
___
uWSGI mailing list
uWSGI@lists.unbit.it
Title: So finally - what is the best way to autoreload uwsgi on file change?
Hello,
I saw here several times questions about uwsgi autoreload, what is useful during the development.
So what is the best way currently, to do autoreload if some files changed/updated?
1. use emperor touch?
2. use
Hello Roberto,
thank you ))
will go the same way. for eclipse PyDEV users this is very handy
thing, as they can just add touch /some/file on every save operation.
so if some file modified, it will be touched :-)
___
uWSGI mailing list
Title: Django in Async mode - any issues to wait for?
hello,
I'am going to use Django in async mode, wrote testing app and seems it works pretty good in async mode, no big performance fault from the prefork mode.
I don't use their ORM, so DB layer must be not issue for me, but maybe there are
Hello Roberto,
I just thought about templates layer, but seems with simple templates
didn't find big difference - about 10% (async 10% slower).
also do you have any idea, why I have these error messages in async
mode?
accept(): Resource temporarily unavailable [utils.c line 525]
accept():
Hello Roberto,
uWSGI 0.9.7.2
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Roberto,
thank you!
you already mentioned about multi fd wait on this list, some time it
can save exec time for sure ;-)
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Roberto,
as you know I'am using multi curl, what is also event based - and
there are own method what detects when data is arrived and it has own
timeout for all operations. it looks like this:
c.setopt(pycurl.TIMEOUT, 10) // timeout for all operations in
multi curl
Hello Roberto,
I have async application, right know I have one issue, what
can be solved with help of threads. all I want it's just to create
20-30 threads what will receive data from the async cores, process
them and return back to async cores.
Do you have any ideas what is the best way to do
)
--
Best regards,
Test157mailto:test...@smail.ee
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Łukasz,
I know and used it, but I want to compare performance just with
by directly execugin code in running script. I know that gevent has so
ability, so I think uwsgi also may have it and asked about it.
___
uWSGI mailing list
Hello Bo,
you are right, I had a mistake with my scripts
1. they didn't return 200 CODE
2. they had print statement to LOG output
so here is test with updated version: http://pastebin.com/nUmW337f
I still don't know though why these results are not stable - server
not overloaded during
I also forgot to mention that I'am using PGBOUNCER in session mode,
statement mode doesn't give much gain in prefork mode.
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Łukasz,
results with PG Bouncer you just saw from me, it doesn't work good because
every time you have to do REAL connection to PGBouncer (with directly
postgresql connections results are much worse), when you are using
pycopg2 connection Pooler - you don't have to do real connections,
sorry previous letter again sent incomplitly, here is full version
Hello Łukasz,
my software has to download 5-7 files on each request. each file is
about 8KB in len but it takes up to 2-3 seconds to download it (time
needed for their generation). so now image, how many processes I have
to fork
Hello Łukasz,
just tested your code: http://pastebin.com/6cymFpsv
[pid: 5446|app: 0|req: 1/1] 10.1.1.10 () {40 vars in 652 bytes} [Wed Dec 22
12:58:28 2010] GET / = generated 2 bytes in 19 msecs (HTTP/1.1 200) 1 headers
in 44 bytes (0 async switches on async core 0)
[pid: 5446|app: 0|req: 2/2]
Hello Łukasz,
here are for comparsion
benchmarks in SYNC (PREFORK) mode:
Postgresql: http://pastebin.com/9c9KBgLm
Mysql: http://pastebin.com/ph7rkScn
SQL ALchemy: http://pastebin.com/LdHU36q7
the same queries, the same amount of req, connections e.t.c.
Hello Andy,
when I last time tried to use gevent, it was unstable under heavy
load, but as I remember I used http://gunicorn.org/ web server, what
is based on gevent
___
uWSGI mailing list
uWSGI@lists.unbit.it
Hello Roberto,
what is your personal email? so I could send you source code.
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Title: Re[2]: [uWSGI] uwsgi django cache backend
Hello Yann,
seems you can't use memcache in async mode atm with uwsgi (hope that builtin cache can work in async mode)
but there are listing of super cool functions in memcached - what uwsgi can't do at the moment, like:
CAS ! (super cool
sorry, here is my full message, those one just was sent incomplete
Hello,
today I did some tests, and found that mysql is about 10 times faster
than postgresql, I also tried to find what wrong with postgresql, so
here my results and I need some suggestions.
at first, how I did benchmarks, if
Hello Riccardo,
sorry :-) yes
thank you for the patch Roberto, with this patch it's compiled ok.
but in in UGREEN mode uwsgi became very slow, so will stay with async
only till the first beta of 0.9.7
___
uWSGI mailing list
uWSGI@lists.unbit.it
Hello Roberto,
well, as my program code became more complex with more async switches and
with more io wait/reads, I got in problem - now sometimes, async requests
can never END. it's more often happened under heavy load, but also
can happen - when almost no load.
I also found, that than more
Hello Roberto,
ohh and once more important thing, harakiri doesn't work. so it
doesn't restart processes, I still has timeouts longer than harakiri
timeout + I don't have any notices about process restart (I have
disabled logs, but as I remember it must notice anyway about harakiri
restart)
so
Hello Roberto,
with each request now I'am getting information like this to console. how to
disable/enable it?
http://pastebin.com/i8r4vusB
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Roberto,
yeah I still have this issue. I tried different linux kernels, and
problem still exist. even on LOW load, I had so connections. Also - on
newer linux kernel I'am getting slower results, and on highger python
version - 2.6 and 2.7 - I have slower speed, than at old Python 2.5
soon
Hello Roberto,
what about some expiring mechanism?
or if all memory is busy, then empty most not used item?
if really both variants, like they are in memcache will be great, and
seems pretty simple to do)
___
uWSGI mailing list
uWSGI@lists.unbit.it
Hello Roberto,
nofile - it's not the reason, user can open upto 32K file pointers
here is some system setting you can check http://pastebin.com/FTXVez2k
usually they allow me to open about 30K connection on nginx with no
problem at all.
here is also my uwsgi cmd: uwsgi --socket
Hello Roberto,
and just forgot to mention that I'am starting to get these messages in
logs from the first second when test started.
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Title: Re[2]: [uWSGI] uWSGI Digest, Vol 15, Issue 9
Hello Lloyd,
you can use apache bench tool, usually it comes with apache package
ab -n 10 -c 500http://yourdjangosite.com/somtestdir/
what means that it will run benchmark with 500 sim. threads, and will do 10 requests.
Title: how thread safe are vars/defs/classes created with post_fork_hook?
Hello,
how safe to share in async mode data in variables created withpost_fork_hook?
how safe so operations for:
1. read only access
2. write only access
3. ready/write access
so can I create one updateable object and
Hello Roberto,
1. if I will create 1000 elements it will pre allocate 1000 * 65K?
2. how good idea to use it for few millions of items? or it's better
to go with memcached then?
3. get/set/del operations are asyn or they are atomic?
___
uWSGI mailing
Title: about 1% of connections getting crashed in async mode
Hello,
When I'am running this code
http://pastebin.com/8ipWTjcn
I'am getting about 1% of connection with this status in logs:
epoll_ctl(): No such file or directory [async.c line 93]
epoll_ctl(): No such file or directory [async.c
Hello Riccardo,
thank you for adding cache.c :)
now works fine
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Roberto,
is it safe to use this code with pos_fork_hook in async mode?
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Roberto,
sorry, found answer by myself. just so code can check connection
status in this way.
# check connection status
try:
c_ok = True
s.getpeername()
except socket.error, why:
c_ok = False
___
uWSGI mailing
Hello Roberto,
I just compiled Python 2.7 - and again, the same issue :-/
very strange
___
uWSGI mailing list
uWSGI@lists.unbit.it
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
Hello Roberto,
thank you for your response Robert,
your version is almost identical to my one, but I took your version now and got
the same results. it's just FREEZE forever. you can check it here:
UNIX SOCKET CLIENT: http://pastebin.com/89LTMzEm
UNIX SOCKET SERVER: http://pastebin.com/jLwb9eQt
Hello Uwsgi,
I slightly modified your async iobound example:
http://projects.unbit.it/uwsgi/browser/tests/iobound_async.py
and got in stuck, here you can see my version: http://pastebin.com/37y0qWEU
instead of TCP/IP it's using UNIX SOCKET + I removed errno checking,
as UNIX socket
anyone has here experience with using async uwsgi with ugreen mysql?
seems the only one driver avaialble I found is
http://github.com/mthurlin/gevent-MySQL
so it's the ONLY way to go if I want to use mysql with ugreen UWSGI?
or maybe there are some better practise?
could you please say, if I'am right or wrong?
if I have so args: --processes 4 --async 1 -l 1024 --master --ugreen
does it mean that uwsgi can accept up to 40.000 clients and can have
additional 1024 clients in backlog?
thank you!
___
uWSGI
60 matches
Mail list logo