Ceph

h1. -1F - Enforced bucket-level quotas in RGW

<pre>

*** mtk has left #ceph-summit1	11:16

scuttlemonkey	wido/gregaf: one of you guys want to take over sage's notetaking duties for this session?	11:16

wido	I have to go afk in a moment	11:16

scuttlemonkey	k	11:17

liwang	hi	11:17

wido	sorry	11:17

scuttlemonkey	li!	11:17

paravoid	wido: don't get me wrong on the varnish-specific stuff	11:17

scuttlemonkey	so great you could make it	11:17

paravoid	wido: I love Varnish and we heavily use it	11:17

wido	paravoid: No problem :)	11:17

paravoid	wido: I just don't think the two layers should be intermixed	11:17

liwang	yes	11:17

wido	No, they shouldn't indeed. I just wanted to point out the caching issues	11:17

liwang	does the invitation sent	11:18

wido	I want to have a "hook" to tell the cache to purge object X	11:18

scuttlemonkey	liwang: in approximately 42 minutes we'll have a break and I'll make sure you get added to the speakers	11:18

wido	And don't do the caching in RGW, since Varnish is much better in HTTP caching	11:18

paravoid	sure, that's not a bad idea	11:18

paravoid	you could invalidate with VCL too	11:18

paravoid	catch PUT/POST and call out to purge (what was previously set obj.ttl=0s)	11:19

liwang	i mean for the bucket-level quota session	11:19

paravoid	thinking Swift here, not sure about S3	11:19

wido	paravoid: True, but if you have multiple Varnish caches, you still get cache inconsistency	11:19

scuttlemonkey	liwang: you want to join as a speaker? Or just as an observer?	11:19

wido	It will only purge locally	11:19

paravoid	indeed	11:20

scuttlemonkey	youtube stream is at: http://youtu.be/rbVFozB9qls	11:20

wido	I'm thinking about having 10 Varnish caches	11:20

liwang	I registered as interested party for this session	11:20

paravoid	we have layered varnishes with consistent hashing among them	11:20

scuttlemonkey	ahhh!	11:20

paravoid	if you don't do that, then you cache the same URLs multiple times	11:20

wido	paravoid: True, you can do very cool stuff with Varnish that way	11:20

paravoid	it's better to shard your cache	11:20

wido	So you have multiple layers of Varnish servers?	11:21

scuttlemonkey	liwang: url sent via private message in irc	11:21

*** paravoid_ has joined #ceph-summit1	11:26

*** paravoid is now known as Guest4609	11:27

*** paravoid_ is now known as paravoid	11:27

paravoid	damn	11:27

paravoid	what did I miss?	11:27

paravoid	21:21 < wido> So you have multiple layers of Varnish servers?	11:27

paravoid	wido: yes	11:27

Guest4609	yes	11:27

*** Guest4609 has quit IRC	11:27

wido	paravoid: Ah, ok. I just tried to avoid that	11:28

wido	So you have them all as each others backend	11:28

wido	Would love to see that VCL :)	11:28

paravoid	not a VCL	11:28

wido	Inline C?	11:28

paravoid	it's a custom director, for consistent hashing	11:28

wido	Ah, that's cool	11:28

paravoid	https://gerrit.wikimedia.org/r/gitweb?p=operations/debs/varnish.git;a=commit;h=885b7d65fb7d93d0b3caa3fbf59a90fc2fffbaff	11:30

paravoid	anyway, this is getting offtopic	11:31

*** Elbandi_ has joined #ceph-summit1	11:32

sergey	I think Google App Engine can do a "503 Over Quota"	11:36

*** ksp has left #ceph-summit1	11:36

wido	paravoid: Aha! Indeed offtopic, but I met one of the Wikimedia people at a VUG in Amsterdam. Got inspired :)	11:39

*** Ryan_Lane has joined #ceph-summit1	11:39

paravoid	that was Mark, the commiter of the above	11:39

mikedawson	sagewk: what is considered a large size Teuthology ceph cluster? What is a long-running test?	11:40

sagewk	mikedawson: other channel :)	11:40

scuttlemonkey	^^ (#ceph-summit2)	11:41

mikedawson	what is considered a large size Teuthology ceph cluster? What is a long-running test?	11:41

mikedawson	sorry	11:41

*** ccourtaut has quit IRC	11:49

scuttlemonkey	new Track 1 video is live on the event page and can also be accessed here: http://youtu.be/OUxEk87RcnI	12:03

</pre>