Web console causes Chromium browser hangs

Greetings,

I’ve observed an issue over the last few months with the station grid in Chrome. If I leave the grid displaying for my station, after 2-3 days, updates stop and the entire browser window becomes useless. I can’t reload the existing tab to make the console work again, I also can’t open a new tab in the same window with the same URL. Both of these result in the browser just spinning while trying to load the page. I have to open a new window to get a working console again. Below is a screenshot of the hung browser, not that I expect it to be all that useful.

Any ideas what the web console is doing that would mess a whole window up? I’ve tested this on various versions of Chrome and several Chromium based browsers, all exhibit this behavior. This is on a Windows 10 Pro machine. Let me know what other info would be useful, as troubleshooting client-side web apps isn’t my thing. :slight_smile:

Thanks!

memory usage when it hangs … might be a memory leak problem from browser …

I will check the memory usage of the browser windows next time it happens. For system memory, I’ve got 32GB, and during regular usage, it hovers around 50% utilization.

It hasn’t hung yet, but I think memory may be a good cause. In the Chrome task manager, I see that tab using nearly 330MB of memory, with almost 200MB of memory used by JavaScript.

It could be the websocket lost connection.

Almost 2 hours later, and total memory used is up 25MB and JS memory is up 20MB.

Unless this is dropping the WS connection a lot, this is creeping up slowly but consistently over the last 2 hours I’ve been watching it.

Another update…

Have you filed a bug report with Google?

What identifiable bug is there to report to Google? From looking at these memory stats, WFs webapp appears to be hanging on to data for too long. Looking at Google’s explanation of the of the Memory footprint and JavaScript Memory columns, an excessive amount of DOM and JS data is building up over time.

Webapp development is not my forte, but I do enough other development to recognize memory leaks, and this feels like an app issue. I’m waiting for a memory snapshot in Chrome to finish processing and see if anything interesting pops out, but before I started that process, I was up to 500+MB of memory, with 350+MB of JS memory being used. For what its worth, I also see this morning I’m getting large CPU spikes every few seconds, which also leads me to believe this is an app issue, as the JS needs to sort through all the data that its holding on to, the rapid wind samples feel like a good cause, considering the timing.

What does the current version of MS Edge do? Since it is also using the chromium engine it would be interesting if it does the same thing.

I do see similar issues with the Opera browser (also Chromium based)

I restarted my Chrome test, and started a Firefox test and the memory usage on both are scaling similarly. I’m going to give this til the next hang in Chrome and see how Firefox is handling.

From my memory snapshot earlier, I see a lot of objects hanging around from the moment library, could be that maybe creating dates from the epochs aren’t getting released properly? The snapshots in Chrome get useless once usage is above 20 or 30 MB, as the tree builder it uses appears to scale really badly as the snapshot gets bigger.

After watching this test for nearly a day, Firefox is suffering the same problem I was seeing in Chrome, and while I change my opinion that it is hung, the JS appears to be so busy, it is not keeping up with WS events in real-time. The X min ago counter is slowly growing larger as it falls further behind. Firefox is suffering the same issue that I can’t open a new tab in the same window to reload the web console, and have to do so in a new window. I’m thinking this is because JS is consuming a 100% of a CPU core and the JS in the new tab can’t get any CPU time, while in a new window, a new JS VM is created and the OS gives it time to run. Chrome is probably running similarly, so I see similar effects.

In both Chrome and Firefox, memory usage has grown consistently over the test period, with both Chrome and Firefox using 150+MB of memory. The presumption here is that event data or other derived objects used internally are being held onto and never released.

WF, is this bug report sufficient to troubleshoot/fix the problem in the web console, and is this thread enough, or should I file a support case and reference this thread?

Thanks!

If you want some action taken it will be faster to tag @corrineb and submit a ticket.

Maybe that is why my Windows RT tablet stops updating the WF data and I have to kill Edge and restart it to get the display working again. It usually runs for a day or two before it appears to hang.

I had the same issue and reported it.
The solution is to open a new tab and paste the your station URL. I sometimes play with cookies and crash the hung tab; this essentially starts a new fresh tab.
The issue stopped when the revised web site was introduced but in the last few days, it has started again.

Thanks for the detailed info on what you are seeing. We will set up a few benchmark tests here and see what we find. One thing that is on our shortlist is to move away from the Moment.js library.

3 Likes