I have a running survey with a lot of traffic and answers coming in. Suddenly, it is not possible to advance past the welcome page in the survey. The welcome page loads fine, but pressing "Next" does nothing. Equally, it is not possible to log into LimeSurvey.

I already set the memory limit of the server to 256M.

What could be the problem here?

Some session issue, since you didn't change anything else : check with your hoster

Thanks for the hint. We are trying to follow up on two leads: our hoster and LimeSurvey in parallel. Unfortunately, we're not hearing back from our hosting provider.

You can try to move to DB session (see config.php file). But since we don' know your hoster …

Suddenly,

This "suddenly" always indicates that something has been changed. Either something has changed in Limesurvey or something has changed in the server's configuration.

However, Limesurvey doesn't simply change on its own. So if you, or anyone who has access to your Limesurvey-Installation did not change anything since the last time everything worked fine, the problem is with the server configuration. Often hoster run updates without informing their users (or the users ignore the information). For me, this is the most probable cause.

But only you know, if and what you have changed in LS.

After our hosting provider finally having reached back out to us, we were able to pinpoint the issue down to caching and running PHP sessions. At the time, we were running a large survey with about 14'000 completes in 16h. On top of that, there was a significant number of people who clicked on the survey link but ended up not completing the survey. We will have to see how we can optimize our process in the future to prevent the LimeSurvey cache from overloading.

Yes, Limesurvey creates fairly big session files and depending on your settings, they might be stored for a while. 14.000 in 16 hours is quite a bit, almost 1000 responses per hour.

Depending on the type and length of the survey, you might shorten the time the sessions are kept alive after the last action. If you use tokens, your respondents can always go back to the survey they started via the token link and they can continue from where they left off (of course, this depends on your settings). Now if it is an open survey without tokens you'll have to find a compromise between session time and the possibility to loose respondents who have stopped responding for a while, but would actually continue later and between those that stopped and won't come back.

And of course, as always, it is a good idea to spread out participation over a couple of days, if possible. With an open survey you might not be able to control this all the time, it all depends on how you 'recruit' your participants.

Thank you so much for taking the time to help get to the bottom of this issue. In the meantime, we have now managed to complete the survey.

The survey was in fact an open survey, the link to which was shared on a website with a lot of traffic. Hence, we were not able to directly control the flow of participants. We will have more surveys liked this in the near future, so we need to do what is possible to reduce the accumulation of session files. What levers are there to tackle this?

You already mentioned session time. I think it is practical to expect participants to go through with the survey in one go without having the possibility to leave the survey open for a long time and then returning to continue. So we can keep the session time quite short. Is that a LimeSurvey setting, or something that we have to set in our server configuration? I have found an age-old forum post on this pointing to the General Settings tab, but this seems to not be relevant in the current LimeSurvey version.

How does survey design affect the session file size? I understand that survey length (number of questions) is important. Do questions contribute to the session file size even if they are not shown to a participant (based on relevance or "always hide" setting)? Are there other dimensions that contribute to session file size?

I am not an expert on this, but first of all I would provide a powerfull machine with a lot of RAM.

Of course the session file is heavily impacted by the number and type of questions within your survey. The more data you collect, the more data is stored in the session file. As far as I know, the whole survey is basically in this session file somehow. But Denis probably can give you more information on this.

And if you have a long session time (I think this needs to be set on a server level), then these data are kept even for those that are long one and won't come back.

So check how long it takes people to do the survey, then check your sessions settings.

From within LS I think there is little you can do, besides keeping your survey short and simple and do not store data that you don't need (e.g. timings, IP, etc.).

Not sure if it helps to switch to database sessions.

gsoto wrote: Is that a LimeSurvey setting, or something that we have to set in our server configuration?

The session lifetime is configured via php settings.
www.php.net/manual/session.configuration...ssion.gc-maxlifetime

The issues you described are typical for a LimeSurvey with a lot of visitors and a server hitting a I/O limit. When a user follows a survey link LimeSurvey triggers the creation of a session file (most common is a session file created in a serverdirectory.

Depeding on the webserver php stack the bottleneck is often not CPU power and RAM, but the storage performance and sometime a storage shortage (depends on where the session files are saved). The sessionsfiles of LimeSurvey are uncommonly big for a php application.
I recommend to use SSD storage with a high I/O throughput. On shared hosting accounts sometimes the session directory is shared with other websites. Which can slow down the sessions operations of PHP when there are a mass of session files. Worst case is many LimeSurvey installations sharing one session directory. There are many things you can tweak and optimize but start with a pure SSD storage.

I talked to our hosting provider yesterday and they told me that session lifetime needs to be set on the application side. However, I'll get back to them with the info you gave me to see what they can do. Thanks a lot!

gsoto wrote: I talked to our hosting provider yesterday and they told me that session lifetime needs to be set on the application side.

Which is no contradiction to what I stated.
The session lifetime is set via PHP. Depending on what kind of webserver / php setup LimeSurvey in running, the php.ini settings can be changed via htacess .user.ini or somewhere else on your side.

LimeSurvey is not offering a setting to override the session setting of PHP.

BTW: Reducing the session time will not prevent the bottleneck you experienced.