722874 - Etherpad page breaks with +10 concurrent users

Status: RESOLVED FIXED

(Websites Graveyard :: etherpad.mozilla.org, defect)

Description

[Matt Thompson]

We use etherpad for our weekly community calls. 
But the pad tends to break once we have a lot of content and more than 10 concurrent users.

Today for example, this pad broke just before the start of our meeting:
https://etherpad.mozilla.org/January

Second time this has happened this month.

Please advise as to:
* whether anything can be done to fix, or if
* we should switch our practice instead. 

Thanks!

Comment 1

[Jake Maul [:jakem]]

How does it break? Like, what happens?

I know there's a hard limit that defaults to (I think) 16 users, but we upped that a while ago to 32 or something- someone ran into it. So I have at least a little reason to believe greater than 10 is workable, but it is admittedly rare enough for us that I would not be too surprised to hear that it has issues.

For testing, you might also want to try making a team site pad and setting it to public. I have no idea if that might behave better or worse, but if nothing else you'll have more control over it. :)

Comment 2

[Matt Thompson]

What typically happens is:
* we get a "restoring connection to the server" message
* the pad tries to reload in the browser, but it just spins endlessly 
* sometimes we get a "server temporarily unavailable" message (not sure if that's the exact right text --sorry I wasn't able to screen grab or reproduce it


* We typically have 30 - 50 people show up for these meetings. 
* I think what I'll do is switch to starting a fresh pad for each meeting. It seems to be more stable when there's less content on the pad.
* Let me know if you have any other suggestions? Thanks!

Comment 3

[Mari Moreshead]

I just had an experience like this but with only 5 people using the pad at the same time.  Not sure if it's the same issue.

Comment 4

[Jake Maul [:jakem]]

@Mari: That's probably unrelated. We had issues this with our PHX1 datacenter (load balancer trouble, as I understand it) this morning from around 8-9am PST... this would have caused similar issues.

Comment 5

[Jake Maul [:jakem]]

We have just updated our Etherpad instance (today)... please let us know if this still occurs.

Comment 6

[Replica @ModReplica]

I'd like to add an update  - the weekly webmaker call etherpads, each one a new instance, are failing each time we use them for our weekly webmaker calls now. The last five calls have had the pads drop offline. 

Can you please describe how to set up a group instance for the communications team and assign a user/attendee level of up to 100?

(eg. the pad that dropped this week for the majority of our call: https://etherpad.mozilla.org/Nov20)

Comment 7

[Jake Maul [:jakem]]

I'm hopeful that this is largely fixed now. Nov27 went off without a hitch, and we've taken preventative measures that should make Etherpad more resistant to the most common failure scenario (which is an out-of-memory condition).

You will probably have trouble beyond 64 concurrent users... I believe that is the limit for a simple public pad like this. If you need more, I recommend you create a "team site", and make the pads within that... the user count on this is effectively unlimited. You can mark team pads as being publicly accessible, to preserve the open nature of these meetings.