We need to load test the banner links for when people click on banners. When a user clicks a banner, the request is directed to the Affiliates websites, a counter is incremented, and then it is redirected to the FF download site. Currently, it redirects to the Mozilla homepage. Here is a working banner link that can be load tested: https://affiliates.allizom.org/banners/link/2/1/1
Assignee: nobody → mbrandt
Whiteboard: [qa+] → [qa]
Created attachment 565008 [details] jmeter script The script can be tweaked to do more concurrent requests as needed.
I believe QA is all set to do some load and soak testing with the JMeter script. I just want to get the go-ahead from ops before doing potentially mean things to our staging env so there are no surprises.
Assignee: mbrandt → cshields
The jmeter script has been run against stage. 1000 concurrent users each submitting 1000 banner GET requests (https://affiliates.allizom.org/banners/link/2/1/1). Stage did noticeably slow down when browsing and account creation but did not come down. No 500's .. no traceback emails. cmore: does this cover the intent of the bug?
Assignee: cshields → chrismore.bugzilla
Status: ASSIGNED → RESOLVED
Last Resolved: 7 years ago
Resolution: --- → FIXED
Matt: did you check if the leaderboard incremented the same amount as the number of jmeter GETs? Also, can you try 10,000 and see what happens? My concern is that 10,000 people won't be using this site at once, but after banners are placed all over the web, the clicks are going to be coming in from X amount of websites with who knows how many eyeballs on them. Thanks!
We will load test production after the password is removed.
Summary: Affiliates: load test banner link on stage → Affiliates: load test banner link on stage/prod
From my machine if I load test production's homepage with 100 concurrent requests that repeat 10 times, they finish in 600-700ms.
Created attachment 568175 [details] Load test on banner link With 100 concurrent requests, the final request only took 700ms to redirect.
With 200 concurrent requests occurring in a forever loop (100 x 2 machines) the response times while browsing prod from a 3rd machine latency was acceptable.
Is it possible to help out with load testing? // Paul Booker
Status: REOPENED → RESOLVED
Last Resolved: 7 years ago → 7 years ago
Resolution: --- → FIXED
Hi Paul, at this point I think we're pretty golden on performance testing. If we do another round of perf testing against Affiliates I'll let you know. Are you interested in perf testing in general? I'm sure other projects could use some help. Bumping to QA verified. We'll reopen the bug or file a new bug if necessary.
Status: RESOLVED → VERIFIED
Product: Firefox Affiliates → Firefox Affiliates Graveyard
You need to log in before you can comment on or make changes to this bug.