Xiaofan Hu

hi @gerases , I'm not 100% sure tbh. do you see any errors in go-carbon logs? I think file list scanning doesn't follow symlinks. it isn't a usual setup that...

(sorry, pressed wrong button when commenting). ticket re-opened.

Hi, I think the the `cache-scan` is still influenced by `scan-frequency`. You might have to lower down `cache-scan` if you want new metrics to be available faster. Also scanning the...

off topic: I think I can add a new feature to [my wip concurrent trie index](https://github.com/go-graphite/go-carbon/pull/334) to make new metrics immediately available, without waiting for scanning being triggered. Should be...

> I suspecting that it works only with trigram index enabled. @deniszh it's a good point, by reading the code, it should works for both index types: https://github.com/go-graphite/go-carbon/blob/master/carbonserver/carbonserver.go#L624-L638 However, it's...

> A delay of 5 minutes (as per scan frequency config) is only 5 minutes. In this environment, the fetchable-after-creates delay is only 7 minutes, but in some of our...

> I also seem to be having a problem, since switching to go-carbon 0.15 and enabling carbonserver and the trie options, where it seems load can temporarily shoot up from...

Have you try running a find command on your production environment? How many metrics do you have there? I think you might an issue on file system or hardware level....

> One thing I'm unclear about; why are the tries required for scanning the filesystem when it's already known that the wsp files dont yet exist on disk? That's the...

I would also suggest running `sudo perf top` to see what the cpu is doing.