| *** eranrom has joined #openstack-storlets | 05:44 | |
| *** eranrom_ has quit IRC | 05:46 | |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: Create an executable file for daemon_factory https://review.openstack.org/348447 | 06:54 |
|---|---|---|
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: WIP: Implement file manager https://review.openstack.org/348482 | 06:59 |
| *** eranrom has quit IRC | 09:04 | |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: WIP: Implement file manager https://review.openstack.org/348482 | 09:19 |
| *** eranrom has joined #openstack-storlets | 09:41 | |
| *** eranrom has quit IRC | 09:43 | |
| *** eranrom has joined #openstack-storlets | 09:46 | |
| *** eranrom has quit IRC | 09:56 | |
| *** eranrom has joined #openstack-storlets | 10:10 | |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: Create an executable file for daemon_factory https://review.openstack.org/348447 | 10:48 |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: Create FileManager class https://review.openstack.org/348482 | 13:59 |
| eranrom | kota: takashi: Hi, I went over the dead lock conversation and wanted to highlight the following point. when a process writes to a pipe write end, and no process reads | 14:00 |
| eranrom | from the read end, then after writing several KBs the writer gets stuck. | 14:00 |
| eranrom | probably some os buffer gets flled. | 14:01 |
| eranrom | s/flled/filled | 14:01 |
| eranrom | So the dead lock was about the middleware being busy writing (in the main thread) while the storlet is also busy writing, and nothing consumes the storlets output | 14:03 |
| eranrom | thus the storlet will get stuck not reading anything from the middleware which ends up getting stuck as well. | 14:04 |
| eranrom | The idea is to do the writing in a different thread so that the moment the storlet writes something we can return from the middleware code letting Swift/WSGI | 14:05 |
| eranrom | consume the app iter. | 14:05 |
| eranrom | I believe that the app iter consumption is done using eventlet non-blocking IO which is in line with doing the writing in an eventlert green thread | 14:07 |
| eranrom | I was not sure if the conclusion from the discussion was that there is a problem there or not. Please let me know. Thanks | 14:09 |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: WIP: Use scope instead of account inside gateway module https://review.openstack.org/348510 | 14:15 |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: WIP: Use scope instead of account inside gateway module https://review.openstack.org/348510 | 14:26 |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: Create an executable file for daemon_factory https://review.openstack.org/348447 | 14:33 |
| openstackgerrit | Merged openstack/storlets: Refactor start_logger https://review.openstack.org/345883 | 17:51 |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: Create an executable file for daemon_factory https://review.openstack.org/348447 | 18:08 |
| openstackgerrit | Takashi Kajinami proposed openstack/storlets: Make daemon_factory return status value https://review.openstack.org/345885 | 18:33 |
| *** openstackgerrit_ has joined #openstack-storlets | 19:35 | |
| *** openstackgerrit_ has quit IRC | 19:36 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!