yoavcodes / highfin Goto Github PK

X-Proxy-Cache, etc.

there's an issue with guppy run fishtank where sometimes the port the salmon app uses will remain blocked when fishtank is relaunched. This only happens on filechange when guppy relaunches fishtank. guppy hasn't released the port back to the os so the salmon app can't bind to it.

Fishtank should keep track of app exec exits in a given time frame.. say 10 failures in 15 seconds will trigger some kind of alert

reproduce: Saving a file too frequently eg: 10 times in 10 seconds, sometimes happens when you make a series of micro edits in a file and hit command+s between each one. If the timing is not too fast and not too slow it can cause guppy to drastically slow down and appear stuck while "Installing..." is printed in the console for every save.

issue: multiple go processes are spun up consuming cpu and ram for each install os.exec. Eventually it will either finish, or those go procs will crash / out of ram causing unexpected behaviour.

solution: If it's in the middle of installing, it should not trigger anymore installs until it finishes.
once the initial install is complete if there's another one waiting it should not run the apps, but rather run one final install so that the latest code is installed and then run the apps.

Fishtank can simply ln the /data folder for a project into instance folders.
aquatic can mount the /data folder via nfs. and use GlusterFS for scaling,striping,sharding,etc. data

some kind of http dashboard, requires project credentials. either over http or ssh. might be able to re-use existing keys.

• maybe guppy connects directly to fishtanks stdout/stderr
• or http page with an iframe that streams fishtank's stdout/err
• fishtank asks for credentials if not already set.
• credentials would be server bound, not project bound.

needs to keep track of:

• stdout/stderr as long
• critical past events. fishtank should maintian a log file containing
• any permanently down apps. ie: 10 failures in 15 seconds we stop trying to restart it
• which apps are reachable via http and their lan response time (separate json response/server handler)

send an email to sysadmin when app exec exits too frequently

skips tests and node.js install checks and just restarts your server on file-change

less compilation only works if you run ./guppy run with a guppy binary in the /vagrant root directory. if you just do guppy run less relative paths are off and can't be found.

all relative paths should be from the -.json directory not the guppy binary dir

when installing a node.js version we don't need some of the npm libraries that guppy needs. need to update the shared node lib to support both use cases.

• weird ssl certificate issue when running vagrant up due to changes in vagrant infrastructure
• need to rebuild the fry-box using latest vagrant