Survey123 Android App - Intermittent Crashing opening one survey

[CoreyLunty1]

Hello

I have a group of users who are experiencing intermittent crashing (the Android 'Survey123 has stopped working' dialog) when opening a survey.

Here are details on the crash:

-They are using Samsung Android tablets (exact model varies, I can get more details)

-Issue has occurred both in office (on WiFi) and in the field (on LTE connection)

-The error only occurs for one survey

-The error has been reproduced on multiple android tablet devices, suggesting it is not a single device specific problem

-The survey is launched by users clicking an arcgis-survey123:// URL, specifying both the itemID and a value for a single field. 

-I was once able to reproduce the problem by starting the survey directly from within the Survey123 app and not via URL

-They typically click this link from within Collector, but I have tested clicking the link from the AGOL webmap in the Google Chrome android app and it also will occasionally result in a crash.  I believe this rules out the transition from Collector app to Survey123 app as the culprit

-In my testing I have reproduced the crash on maybe 3% of survey launches.  Field staff reports suggest they are seeing higher failure rates however 

Here are resolutions we have attempted:

-Deletion and redownloading of the problematic survey via Survey123 app

-In Survey123, going into Settings > Storage > Fix Database

-Uninstall and reinstall of Survey123 app via Google Play store

The problem may not happen again for a while but it always reoccurs eventually.

Here are some things I would rather not do:

-factory reset tablets

-have to republish the survey

Note the large log file involved me repeatedly attempting to reproduce the problem. I have included the whole log because of the information at the beginning on android version, app version etc. The other two logs I have clipped to only include logging from the attempts which actually generated an error.

update: removed first and last names of inspectors from xlsx and re-attached esriinfo.zip

[Anonymous User]

Hi Corey,

What version of the Survey123 app are you using on Android, does the same version crash on all devices? Did previosu versions crash with same survey?

Out of curiosity have you tested the 3.4 beta version available on EAC here: https://earlyadopter.esri.com/project/version/default.html?cap=e69ef91f45744b98882c651f7b518eb7&artt...

I see you are using quite a few pulldata() calculations and external csv files, as well as external choices. We have seen with other users that some out of memory issues caused by pulldata() calculations and external csv files have been resolved in 3.4 beta.

Phil.

[CoreyLunty1]

I strongly believe we are currently using S123 v 3.3.64 on all Android devices, though I would need to confirm.

You are correct that I make somewhat extensive usage of pulldata and I had suspected it was this complexity that could be behind the crashing.  We are using an automated nightly process to update those external csv files, and republish the survey via FME Server and Python; they are vital to the workflow.

I installed the 3.4 beta on my testing device and have found it more stable and better performing, and that it has reduced failure rate to maybe 1% from the 3% I was seeing before. However, the field staff were always seeing higher failure than I was, and I plan to get the beta apk installed on some of their tablets over the next couple days to see if it resolves, or at least reduces, instances of the application crashing. 

To help with troubleshooting I installed a lightweight RAM monitor app that displays a screen overlay and if I repeatedly try to start the survey over and over again, the available RAM does start to creep downwards, and then when it gets to around 6% the app crashes, suggesting it is still running out of memory, but the new version seems to better manage memory when using pulldata.  I will find out if the field staff are running other apps that may be RAM heavy - keeping Chrome open with many tabs open or something perhaps.

[DougBrowning]

I wonder if this is the same issue I had.  I found that iOS will only let me hide (using a relevant) a few big drop down lists (via pulldata) before it starts crashing.  

https://community.esri.com/thread/232390-survey123-crashes-on-ios-devices 

[CoreyLunty1]

A single field user was provided the 3.4 beta and performed testing yesterday. They did not experience the app crashing due to running out of memory.  More field testing will be performed today and over the next few days by multiple users and I will be sure to report my results.

Phil, I don't suppose you can share any timeline on the official release of 3.4 in the Google Play store? I'm guessing not, but we are left with a decision to make here regarding either using the beta version on a large number of devices (which some folks are nervous about) vs waiting for the official 3.4 release. Thank you for any info you can share.

[JamesTedrick]

Hi Corey,

We expect the next release of the Survey123 field app in late June.

[DougBrowning]

Just had a big field training day.  Several users ran into memory errors I think.  In this case we have a large repeat that we set to a repeat_count of 50.  Randomly it would blank one of my main calc fields - which breaks my entire form.  The user is then stuck as they cannot go forward or backward.  It is usually between repeat 11 and 15.  Same tablet may be fine for a few runs then dies.  We also see some relevant fields never show up when this happens.  These are Samsungs but my iOS dies sometimes also.

It did seem to be when the user had other apps open, but not always.  It also seemed to happen only on tablets that had the Mass360 management software on them - but this is not confirmed.  So the guess is memory issue.  For now we are taking out the repeat_count in the hope that a user could delete the bad repeat and try again.  Crew is very nervous about this as they may be out in the field for a week at a time.  This is also an issue on 3.4.

We also are trying to teach users to get around the repeat wrong error issue I posted about earlier.  Basically our last repeat has some questions that do not appear on repeat 49.  If a user is on 50 it tells them they are missing a required - even though they are not.  Go to 49 and then it submits.  So far this bug is still in 3.4.   Big bug.  And of course the required issues bug.  I now have 4 major bugs open in 3.4.

I am rather disappointed that 3.4 is now pushed back again to June!  Letting the major bugs of 3.3 go this long, and past the standard field season kickoff, is really not cool.  If anything roll back to 3.2 at this point (can we get it).  You cannot leave core bugs out there that long.  I am personally taking a LOT of heat for this.

Please in the future consider lining up new releases with field season.  Also more testing please.  We have a crew of hundreds siting and waiting on this!  If there is a bug like this it could cause a crew to stay late - and maybe get hurt out there!  Again please consider how some of these things affect us and how many people are on us about it.  The 123 team has been awesome the last 3/4 years but this one was a big opps.

Thanks

[JamesTedrick]

Hi Doug,

Thank you for your feedback.  One of the reasons that the schedule for the next release if the field app has slipped to address the issues that you have mentioned and give enough time to test and validate the fixes. We are working to get the release completed as soon as is practical.

[CoreyLunty1]

Hello James and Phil

We continued testing with the beta of 3.4. Some users continued to have issues with the app running out of memory and crashing. I will be revising the attachments on my original post to include a log file from June which was gathered using 3.4.  

I notice that 3.5 was released on Google Play last week.  Does it contain additional fixes that may help with this 'out of memory' issue?

[JamesTedrick]

Hi Corey,

As you note, 3.5 has been released.  As part of the release, Survey123 is now available in a 64-bit version for Android; this does improve performance on Android devices.