Increase of timeouts when using userId modules, since 6.20.0

[JulieLorin]

Type of issue

This is a bug/regression

Description

We have noticed that we have an increase of timeouts when using a userID module (like criteoIdSystem), since 6.20.0 I think this is linked to : https://github.com/prebid/Prebid.js/pull/8201/files It was working fine on 6.19.0 What we consider a timeout the event CONSTANTS.EVENTS.BID_TIMEOUT

Steps to reproduce

The example is made with Criteo but it seems to be global for user ID module. We have the same issue with ID5 for example. Building with userId + criteoIdSystem is fine, as long as we do not add criteo in pbjsconfig.userSync.userIds. We go from 3-4% timeout rate to 10% just by enabling criteo user sync in config.userSync.userIds

user Sync settings :

userSync: {
  userIds: [{
    name: 'criteo',
    storage: {
      type: 'html5',
      name: 'criteoid',
      expires: 365
    }
  }],
  syncDelay: 1000
},

All browsers are affected Environment with and without GDPR consent are affected We have tried to upgrade to Prebid 7.8.0 to see if things gets better but it does not (no change). I also tried to set the auctionDelay to 1, but this didn't have any impact

[patmmccann]

we (cafemedia) noticed as well, our hope was https://github.com/prebid/Prebid.js/pull/8626/files#diff-381b65d4141bd216f9021175532d712ce87a9c4a55985c0cf1226f03421b6014 or some new combination of settings fixed, but we haven't tested 7.8 or those various combinations yet

[patmmccann]

@jlustig11 confirms Rubicon DM observed as well

[JulieLorin]

Thanks, it's good to confirm it's not just us.

I also tried to twitch the auctionDelay parameter to check if it could improve performances (but it did not). Let me know if you need a second set of hands to test/check things, I am available to help with that if there's anything I can do.

[dgirardi]

@JulieLorin (and also @patmmccann ), are you able to see if using cookies vs local storage makes any difference?

[dgirardi]

I also realized that there might be multiple issues going on - 6.20.0 had this bug which would conceivably explain the increase in timeouts. That was fixed in 6.23.0. Is that also something you could try?

[JulieLorin]

I also realized that there might be multiple issues going on - 6.20.0 had this bug which would conceivably explain the increase in timeouts. That was fixed in 6.23.0. Is that also something you could try?

I tried upgrading to latest Prebid version (7.8.0 at the time), so this fix was present, but didn't change anything

[dgirardi]

@JulieLorin - are you calling any of the userId apis directly (getUserIds refreshUserIds etc), or just requestBids? Would it be possible to set up a page with this issue for me to look at?

[bretg]

Magnite data does not show a rise in bidder timeouts. We upgraded from 6.11 to 6.24 on June 30th. I checked 3 accounts in different regions and did not see any appreciable change in bidder timeout rate.

[patmmccann]

@jlustig11 could you add some detail about the nature of the performance issue you have observed?

[JulieLorin]

@JulieLorin - are you calling any of the userId apis directly (getUserIds refreshUserIds etc), or just requestBids? Would it be possible to set up a page with this issue for me to look at?

We are not calling any of those methods directly (only requestBids). You can use this page on which the issue is present : https://www.parents.fr/ (If you need to enable prebid logging, please use #hbv-pbjs in the URL or localStorage.setItem('hbv:pbjs', 'true') in the console instead of the usual ?pbjs_debug=true)

We have just updated to prebid.js 7.11.0 but no changes have been noticed.

[JulieLorin]

The timeouts returned to their normal value since (at least) 7.17.0 (I didn't test versions between 7.11.0 et 7.17.0), so I think this issue should be closed

[patmmccann]

Phew!