Quantcast
Channel: ClearScript
Viewing all articles
Browse latest Browse all 2297

Commented Unassigned: Access Violation Exception in v8-x64.dll [115]

0
0
Hi!
We launched a new site last week that uses ClearScript to server render a React application. After launch, we get around 2-3 crashes on Access Violation Exception in w3wp.exe a day per web server. The site has around 200 concurrent users on average, with peaks up over a thousand. We use the latest version of ClearScript (5.4.7).

We see that the crash is an Access Violation Exception in v8-x64.dll and we have a few dump files from such crashes, but since we don't have any debugging symbols the stack traces doesn't give us much.

Would you be interested in looking at the crash dumps to help us figure this out? Or if you can help us with debug symbols so we can get more information out of the dumps.

Some information about how we use ClearScript:
1. We create multiple V8Runtime and V8ScriptEngine and keep them in a queue. When a request comes in, we dequeue an engine, run JS and return it back to the queue.
2. Since we don't store any JS globals we reuse the same V8 context for different visitors. But we run CollectGarbage(true) for every 30 execution. We have deployed a change this morning where we call CollectGarbage(false) instead but twice as often. Not because we think it will solve it, but see if the behavior changes.
3. We don't pass any CLR references to V8. The only thing we do is call Engine.Execute with strings of JS code. When we pass data to V8 we do that by either calling Engine.Execute or Engine.Evaluate with strings of JSON.

Any help or pointers are appreciated! We'd be happy to share code with you if you need it.
Comments: Hi! We'll try go get symbols deployed to get a better error description. We have a queue of max 20 runner instances where a runner is one script engine instance and one runtime instance. We never dispose them, we only run GC on them. When a request comes in for a runner but all 20 instances are busy the caller is put in a waiting state until a runner is free. There seems to be some correlation between crashes and used memory. It seems that the crashes occurs most frequently when memory usage levels out at around 60-80%. After a restart we see that memory usage raises (as expected) as caches etc are getting filled. Then memory levels out as cache eviction etc starts to kick in. And that's when we see most of our crashes.

Viewing all articles
Browse latest Browse all 2297

Latest Images

Trending Articles





Latest Images