Closed src7 closed 5 months ago
I am feeding more then 150 pastes a minute....
On Sun, Feb 23, 2020 at 5:41 PM src7 notifications@github.com wrote:
Hello,
why can not we feed more than 150 pastes per minute ? It does not look like a hardware bottleneck.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/CIRCL/AIL-framework/issues/477?email_source=notifications&email_token=AAZXPEOPN6IUMVOFG6R5JDTREL3QVA5CNFSM4KZ6OZ52YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IPS7GQQ, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZXPENM3IDRQXIB2HM5TELREL3QVANCNFSM4KZ6OZ5Q .
Thank you, we tried on different servers and Processed pastes
is stuck at about 150 (on one single feeder, using bin/import_dir.py
)
I will reach out to my team, and see if they had to make any special changes.
On Sun, Feb 23, 2020 at 5:49 PM src7 notifications@github.com wrote:
Thank you, we tried on different servers and Processed pastes is stuck at about 150 (on one single feeder, using bin/import_dir.py)
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/CIRCL/AIL-framework/issues/477?email_source=notifications&email_token=AAZXPEJVSRMAAO6NITYCNH3REL4PLA5CNFSM4KZ6OZ52YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMWJWMY#issuecomment-590125875, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZXPEIGQPGO2SYMLR5QNULREL4PLANCNFSM4KZ6OZ5Q .
Thank you, currently I am trying something with configs/6382.conf
Hi @src7 ! Do you have any stuck queues ? Are you processing huge files ?
It might be a disk issue. We create the Global
module to save all items on disk.
Can you try to launch multiple Global
modules ?:
Screen -r Script_AIL
Crtl+a c
. ./AILENV/bin/activate
cd bin
./Global.py
Hi,
Queues are not empty, but not stuck. Files processed are regular pastes from pastebin.
The disk is NVME or RAM based (direct I/O disabled for the last one).
I will try these commands.
The weird thing is that it is the same constant rate all the time. Could it be one sleep somewhere ? (didn't found it yet)
We compile the last version of AIL directly from the master branch on a fresh install of Ubuntu Bionic.
Imported pastes are not compressed (if it can help)
@src7 Are you using the import_dir.py
script by any chance?
If yes, you might want to change the sleep time to something smaller. Look at this:
https://github.com/CIRCL/AIL-framework/blob/5ae22ec2168a55e89279228de7c4cdbbd36baa44/bin/import_dir.py#L114
Yes, and that is not the problem. I see way more than 150 pastes per minute imported in the console (no sleep set).
Global
, Mixer
or import_dir? By default
args.seconds`` is set to 0.2simport_dir
(also tested with different sleep values and even without the sleep line)Processed pastes
After some research I have found that tuning this delay changes everything :
Currently reaching 4000 pastes per minute and working on more.
Good catch ! I did some modification on the ZMQ feeders. 998f8cc8e15f81dff5a4d006509d5c58883da629 The feeder use ZMQ Poller for more general non-blocking I/O.
You should reach more than 4000 pastes per minute
Hello,
why can't we feed more than 150 pastes per minute ? It doesn't look like a hardware bottleneck.