Comparison to Browserless #398
Replies: 1 comment 2 replies
-
Hello @sagikazarmark👋 I’m not familiar with Browserless so I cannot make a 1:1 comparison 😅 but I’ll try to give you some insights about how you may scale Gotenberg. First, Gotenberg will start a dedicated instance of Chrome per request. Once a conversion is done, the instance is killed to mitigate memory leaking. But it also means that a lot of concurrent requests to a single Gotenberg instance will spawn a lot of Chrome instances. Nevertheless, vertical scaling works out of the box. From my experience, CPU/memory usage will increase in such a scenario, but not exponentially though. Memory usage is also more important than CPU, but the more CPU is available, the quickest the conversions. In order to scale horizontally your number of Gotenberg instances, you have two metrics:
I don’t think memory usage is a « good » metric here because it decreases somewhat slowly over time. I don’t know the reason, because even if there are no more Chrome processes, the memory does not free up immediately (the root cause is maybe Debian?). The second metric is more interesting IMO, but it requires some fine tuning on your side to determine what’s the « best » number of active Chrome instances before scaling up/down Gotenberg. It also requires Prometheus 😅 |
Beta Was this translation helpful? Give feedback.
-
Hey there!
Has anyone compared Gotenberg to Browserless Chrome? At least the features that are comparable.
I'm particularly interested in Gotenberg's scalability options (auto-scaling based on load) and whether there are any performance spikes when using Chrome based feature (there probably are) and how they are handled compared to Browserless.
I spent some time on getting Browserless running on Kubernetes and faced the following issues:
Any insights would be much appreciated. Thanks!
Beta Was this translation helpful? Give feedback.
All reactions