Welcome to the LimeSurvey Community Forum

Ask the community, share ideas, and connect with other LimeSurvey users!

Memory leak in LS 5.4.10

  • David1231
  • David1231's Avatar Topic Author
  • Offline
  • New Member
  • New Member
More
1 year 5 months ago #233890 by David1231
Memory leak in LS 5.4.10 was created by David1231
Hello,
I performed an Apache Bench test, making 100k requests to LimeSurvey 5.4.10. This fully consumes the available 64GB of RAM. Three hours after the test is completed, this RAM is not released/cleaned. Is it possible that there is a memory leak caused by LS?
Thanks!

Please Log in to join the conversation.

More
1 year 5 months ago #233892 by jelo
Replied by jelo on topic Memory leak in LS 5.4.10
That is not unusual for an Linux/Posix server.
You might check with the wrong tools if the memory is really occupied.
LimeSurvey itself cannot have an memory leak. It would be PHP and the handler.

You will see nearly te same memory usage when you use a static html file and run ab with 100k.

The meaning of the word "stable" for users
www.limesurvey.org/forum/development/117...ord-stable-for-users

Please Log in to join the conversation.

  • David1231
  • David1231's Avatar Topic Author
  • Offline
  • New Member
  • New Member
More
1 year 5 months ago #233895 by David1231
Replied by David1231 on topic Memory leak in LS 5.4.10
Thanks for your reply. Indeed, I checked it further and found out that this memory is released from RAM but stored as "cache". Perhaps this explains increased disk utilization. How could I find out what causes this high consumption?

Please Log in to join the conversation.

More
1 year 5 months ago #233901 by jelo
Replied by jelo on topic Memory leak in LS 5.4.10

How could I find out what causes this high consumption?
 
If you open 100k requests on your server what do you expect? The Linux system is trying to use every resources to cope with that.
I recommend to find a forum around your OS or webserver administration to ask these kind of questions.
Take a look at the sysstat package.

The meaning of the word "stable" for users
www.limesurvey.org/forum/development/117...ord-stable-for-users

Please Log in to join the conversation.

  • holch
  • holch's Avatar
  • Offline
  • LimeSurvey Community Team
  • LimeSurvey Community Team
More
1 year 5 months ago #233915 by holch
Replied by holch on topic Memory leak in LS 5.4.10
Now this doesn't answer your question but I wanted to add: getting 100k concurrent requests to a survey usually (I say usually because there might be cases where things are just different) indicates that the strategy of publishing the survey is flawed. Invites are usually send out in waves to avoid extreme peaks at the start of a survey. Any survey system (and especially the underlying server configuration) will have its limits at some point.

How many respondents do you realistically expect? Do they all need to gain access at the same time or can you spread out the invitations over the day, week, month?

In market research we usually never send out all invites at the same time for a couple of reasons:
- to avoid extreme peaks in access (people are generally curious and once they receive the invite, they'll open the link, even if they don't plan on participate in the survey, they just want to see what it is)
- to be able to "correct" things after a soft launch. Often we just send out a few invites to see if everything works well, have a quick check at the database to see if everything is registered correctly.
- To avoid "bad times" for sending out invites. There are always 'better' times with higher response rates, but this highly depends on your target group, so often you don't know when the best time is to invite people.
- etc (I am sure there are more good reasons, but that is what came to mind now.)

I answer at the LimeSurvey forum in my spare time, I'm not a LimeSurvey GmbH employee.
No support via private message.

Please Log in to join the conversation.

Lime-years ahead

Online-surveys for every purse and purpose