Hey all I am trying to figure out why the same django app ru

Question

Hey all I am trying to figure out why the same django app running in docker is almost using double the memory when running using pants compared to running my old manual docker file I am using the template <https github com pantsbuild example django|example django> in my pants project and I have gunicorn configured with same params in both cases ```workers = 1 threads = 8 timeout = 0```

glamorous-nail-59544 · Answer

Dockerfile old dockerfile

glamorous-nail-59544 · Answer

Dockerfile pants dockerfile

glamorous-nail-59544 · Answer

are there any gotchas with `pex` files

happy-kitchen-89482 · Answer

How are you building the pex Have you tried building a venv mode pex

glamorous-nail-59544 · Answer

Almost identical to the example django repo

happy-kitchen-89482 · Answer

< bitter ability 32190> you looked at pex in docker performance extensively any thoughts

happy-kitchen-89482 · Answer

< glamorous nail 59544> have you read for ideas

glamorous-nail-59544 · Answer

no will check

glamorous-nail-59544 · Answer

Actually also startup time ~almost~ more than doubled This the same image deployed to cloud run

happy-kitchen-89482 · Answer

I don t anything about cloud run is that running containers for you

glamorous-nail-59544 · Answer

Yeah just horz scalable and ephemeral container runner Similar to cloud functions v2 EDIT It optionally scales to 0

happy-kitchen-89482 · Answer

Here is some info <https pex readthedocs io en v2 1 120 recipes html highlight=layout pex app in a container>

happy-kitchen-89482 · Answer

And that blog post I linked to above has examples on how to build and deploy venv pexes

happy-kitchen-89482 · Answer

So follow along with that and let us know here how that goes

bitter-ability-32190 · Answer

We never looked at memory Using a venv PEX already pre installed should yield EXACTLY the same results as just making the venv yourself assuming everything else is similar Code l packages etc

glamorous-nail-59544 · Answer

I doing an experiment comparing all combinations will report here

glamorous-nail-59544 · Answer

Need to figure out startup time next to decide the best version For now looks like `venv zipapp` is the middle ground between size and memory

enough-analyst-54434 · Answer

< glamorous nail 59544> what Josh said is true and so it looks like you must not be following the link Benjy pointed to or Josh s blogs Those both have you pre build the venv in the container at container build time That 1 time container build step should make packed loose zipapp all the same in terms of mem and timings

enough-analyst-54434 · Answer

Perhaps this is hard to do with Pants docker integration

enough-analyst-54434 · Answer

I don t know much about our Docker integration I write manual docker files and build them outside of Pants

glamorous-nail-59544 · Answer

I looked at the blog haven t checked pex docs shared above yet Will look at that now Was comparing the available built in options Might get away without the added complexity of the solution mentioned in pex docs

enough-analyst-54434 · Answer

Ok Just know that the tradeoff is a few more lines of Dockerfile for lowest latency + memory usage all in one

bitter-ability-32190 · Answer

With the blog post it s even more lines plumbing but focuses on incremental caches making it worth it

bitter-ability-32190 · Answer

< enough analyst 54434> our docker integration currently uses hand written docker files So complete flexibility

glamorous-nail-59544 · Answer

I missed the blog post above was looking at <https blog pantsbuild org pants pex and docker |this> Working on getting the most optimal version now relieved

glamorous-nail-59544 · Answer

had some fun with detailed benchmarks

curved-farmer-66180 · Answer

< glamorous nail 59544> Hi Can I ask what version did you choose We are using dockerfile optimal multi stage 2pex but still has some memory issue and I am looking into it

glamorous-nail-59544 · Answer

Untitled dockerfile

glamorous-nail-59544 · Answer

< curved farmer 66180> hey I ended up using it as well here is a sample dockerfile

curved-farmer-66180 · Answer

thanks < glamorous nail 59544> It is almost same as ours I found the memory issue was from gunicorn configuration