While using the workspace lately, I’ve noticed a large spike in disconnections, bad LSP performance, constant machine downgrades to 500MB of resources while only one Repl is open, and Ghostwriter’s Complete Code feature hanging for considerable amount of time after I make changes to my code.
I’m on the Pro plan and have observed this behavior in both the Desktop App and the web version.
To further provide more details on my experience, here are the four major pain points I’ve had to deal with for the past ~1.5 weeks or so:
Resource Downgrade - By far, resources going from 4 Gigs of RAM and CPU (Very Fast Workspace provided by the Pro Plan) to 0.5 Gigs of each has been a major issue. This happens quite frequently when I enter a Repl or have been coding for a while (hard to get an estimate, happens pretty randomly)
After my resources plummet, LSP becomes super slow and sometime stops working alltogether. There is no button to reboot/restart the LSP, so I have to end up running a kill 1and refreshing the page to get it working again.
LSP Latency - LSP Has been performing considerably slow, even when my Repl’s resources are 4 Gigs while I’m coding. Unused variables take approximately 2-7 seconds to appear after deleting unused code, code actions also takes a considerable amount of time, and autocompletions don’t appear as often.
Disconnections - It happens randomly. I could be coding or having the workspace open, idle for a few minutes and all of a sudden LSP has stopped working and my Repl has stopped.
Recently, Replit announced that all users, including free plan users would have non-preemptible VMs in the workspace to prevent constant disconnections, which could be loosely related. Around this time, I noticed considerable amounts of disconnections in the workspace.
I imagine that the global amount of workspace disconnections may have dropped by a huge amount, but I’m pretty sure if the analytics on Pro/Hacker users with the Fast Workspace perk were zoomed in and inspected, the disconnection count may have gone up.
Ghostwriter Performance - To be fair, this report might not be a reality since I’ve been constantly jumping around my code and documentation really fast (been using tailwind so I have to check the docs a lot). It might just be me, but I’ve been seeing a bunch less ghostwriter suggestions (complete code feature) lately.
Lastly, providing as much relevant information as possible:
Hello everyone. I have replit hacker that will expire the first of September, by which my workspace should have 2vCPUs and 2GB of RAM. For some reason, sometimes, my workspace will randomly switch from the 2vCPUs to 0.5vCPUs and from 2GB to 0.5GB of RAM.
Problem description:
when I boot up the repl initially after not using it for a period time (like a few hours), i found it has the free tier limits imposed on it , but resolves after refreshing the page and i have my pro allowance fully available. I have ghostwriter available, just the vcpu and ram is capped to free tier limits at initial boot up in session. might just be the provisioning of resources taking time. but i just noticed this happening recently.
Expected behavior:
Pro Plan tier Limits available
Actual behavior:
vcpu and ram capped at Free Tier Limits.
Steps to reproduce:
wait for a period of time of repl being inactive and opening ide hours later.
I am on a pro plan. When I run some repl, it uses 4vCPU and 4gb ram as indicated in pro plan. But then it suddenly goes to 0.5vCPU and 512mb ram (which is free plan)
While using the workspace lately, I’ve noticed a large spike in disconnections, bad LSP performance, constant machine downgrades to 500MB of resources while only one Repl is open, and Ghostwriter’s Complete Code feature hanging for considerable amount of time after I make changes to my code.
I’m on the Pro plan and have observed this behavior in both the Desktop App and the web version.
To further provide more details on my experience, here are the four major pain points I’ve had to deal with for the past ~1.5 weeks or so:
Resource Downgrade - By far, resources going from 4 Gigs of RAM and CPU (Very Fast Workspace provided by the Pro Plan) to 0.5 Gigs of each has been a major issue. This happens quite frequently when I enter a Repl or have been coding for a while (hard to get an estimate, happens pretty randomly)
After my resources plummet, LSP becomes super slow and sometime stops working alltogether. There is no button to reboot/restart the LSP, so I have to end up running a kill 1and refreshing the page to get it working again.
LSP Latency - LSP Has been performing considerably slow, even when my Repl’s resources are 4 Gigs while I’m coding. Unused variables take approximately 2-7 seconds to appear after deleting unused code, code actions also takes a considerable amount of time, and autocompletions don’t appear as often.
Disconnections - It happens randomly. I could be coding or having the workspace open, idle for a few minutes and all of a sudden LSP has stopped working and my Repl has stopped.
Recently, Replit announced that all users, including free plan users would have non-preemptible VMs in the workspace to prevent constant disconnections, which could be loosely related. Around this time, I noticed considerable amounts of disconnections in the workspace.
I imagine that the global amount of workspace disconnections may have dropped by a huge amount, but I’m pretty sure if the analytics on Pro/Hacker users with the Fast Workspace perk were zoomed in and inspected, the disconnection count may have gone up.
Ghostwriter Performance - To be fair, this report might not be a reality since I’ve been constantly jumping around my code and documentation really fast (been using tailwind so I have to check the docs a lot). It might just be me, but I’ve been seeing a bunch less ghostwriter suggestions (complete code feature) lately.
Lastly, providing as much relevant information as possible:
I’m having this issue as well. It seems that anytime I do something that uses the resources I have paid for, I wind up seeing the resources downgrade. At first I thought maybe the resources were only available on one repl at a time. I tried making sure that only one repl was running, but even with one I see random downgrades of service.
Turning off key features, such as language intelligence, is not a fix. Perhaps what we are seeing is that replit does not really have the resources that it is selling. Having private projects is nice, but I really want to make use of the CPU time and RAM that I have paid for.
At the most, you need to close and re-open the tab, if that’s better