796789 - [meta] Calendar Perf and Memory Optimization to Prevent Sync Crashes

Status: RESOLVED INCOMPLETE

(Firefox OS Graveyard :: Gaia::Calendar, defect)

Description

[GH to BZ]

[GitHub issue by dmandelin on 2012-09-28T21:35:33Z, https://github.com/mozilla-b2g/gaia/issues/5459]
Twice today, using the 9/28 build, when I tried to sync the calendar manually (touching the circle-arrows button), this happened:

- button goes away
- time passes...
- calendar crashes

On the third try, it didn't crash, although I also didn't see any events.

Comment 1

[GH to BZ]

[GitHub comment by lightsofapollo on 2012-09-28T22:04:26Z]
@dmandelin Can I have some details on the calendar you where trying to sync?

Right now we do an _entire_ collection sync so its possible we run out of memory before we get to the records in the present (assuming you have a ton of events).

1. what kind of provider? (Google, Yahoo, Zimbra)
2. some rough of idea of # of events ( I had x calendar for 90 years )
etc..

Thanks!

Comment 2

[GH to BZ]

[GitHub comment by dmandelin on 2012-09-28T22:49:20Z]
@lightsofapollo 

OOM is quite possible for me. I'm using our Zimbra. It's probably 2-3 years of events at this point.

Comment 3

[GH to BZ]

[GitHub comment by nhirata on 2012-09-28T23:21:33Z]
I can't reproduce the crash, but then again I don't have 2-3 years of events in my calendar for zimbra.  I think it is a OOM issue?

Comment 4

[GH to BZ]

[GitHub comment by lightsofapollo on 2012-09-28T23:25:52Z]
@nhirata Highly likely, over the next day or two this will be improved as we will only sync last 30 days + future and it should continue to get better. I was able to sync a good 6 months events & recurring events (with no other apps open) but while it only took around 10 seconds on my desktop it was a good 30-60s on my Oroto.

Comment 5

[GH to BZ]

[GitHub comment by jds2501 on 2012-09-29T02:11:55Z]
@lightsofapollo That will help improve the issue being hit here, although I might suggest filing a bug in the platform as well for the crash specifically, as the crash portion of this bug is a gecko problem.

We can track a fix for as you've said for what can be done on the Gaia side to mitigate this.

Comment 6

[James Lal [:lightsofapollo]]

A significant amount of progress has been made to reduce sync times but there is still more to be done. I don't think it will crash since we limit the sync range quite a bit now. Its possible that this still could oom without (796743) for very large calendars with hundreds of recurring events.

Marked resolved for now, please give it another try.

Comment 7

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

The memory usage of the calendar goes up wildly while trying to sync my zimbra calendar.  It reliably crashes within about 5-10 seconds.  That means the rate of memory allocation is up to 10MB/sec.  That is, ah, rather a lot :).

Comment 8

[James Lal [:lightsofapollo]]

P1=top of my queue, but this issue in itself is more a meta bug. 
The calendar crash is due to OOM and there are potential areas for optimization at each layer. I am marking those bugs as blocking this one and will proceed with work on the individual pieces.

Comment 9

[Alex Keybl [:akeybl]]

Milestoning for C2 (deadline of 12/10), as this meets the criteria of "remaining P1 bugs not already milestoned for C1".

Comment 10

[Jason Smith [:jsmith]]

When I was testing the patch from bug 805366 before it landed, I didn't see any crashes noticeable anymore. I'm inclined to believe that bug 805366 might fix this bug.

If anyone is still crashing after you try out the patch from bug 805366, please comment in the bug indicating so.

Comment 11

[Jason Smith [:jsmith]]

Actually wait a second - didn't realize there's one bug left.

Comment 12

[Jason Smith [:jsmith]]

Given that this is a meta bug in reality, I'm changing the bug context to reflect that. The sub-bugs are the actual work items here.

Removing basecamp+ given this is more of a meta bug now.

Comment 13

[James Lal [:lightsofapollo]]

There will always be more we can do, lets open up new bugs or another meta-bug as we need them