Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Current Delivery Times

This site may earn commission on affiliate links.
#6,021
My order changed to the damn invisibility paint again! I know Elon wants to be on the cutting edge, but it's going to make finding it in a parking lot really hard.
Screen Shot 2021-05-20 at 6.09.32 PM.png
 
  • Like
  • Funny
Reactions: MrsV and james-007
#6,028
Did a little digging to figure out where the tesla site is running. Turns out it seems to be hosted on a front end IP address that is owned directly by them. The ASN for the IP that tesla.com resolves to matches tesla as well so this is advertised by them direct to the internet. It does traceroute through akamai, so there's some CDN in front of it for volumetric protection or WAF, but it really seems based on a dig/traceroute/whois/ASN lookup that this site is hosted from their premises, not in a cloud. It could be proxying traffic via load balancers/api gateways from web services in a cloud, but this could explain capacity issues, etc, like a bunch of data center appliances that are melting (WAFs/firewalls that are maxed out on flows/etc), or capacity in a blade chassis/hw cluster that normally can handle it. It wouldn't surprise me either if they were querying an ERP that manages the assembly line for the status of each order, and that ERP has an API query limit that is exhausted by all of us refreshing all day long. I'd put money on the fact that at least some of the outages are failed code pushes trying to figure out how to do a cache layer on that API, or something to offload that traffic. Doubt it's the same crew designing the car facing stuff. Although, I recall a reddit thread from a few years ago that talked about some basement duct tape/bubblegum infra that used to run things. I've seen other stuff saying they've moved to Kubernetes, etc for the car facing apps, but this account site could be a lot of tech debt that is now coming due for payment.

**I'm a cloud nerd who doesn't want to write documents for work right now, so I did this instead. Not making excuses for them (at least just put up a page that throttles requests/static caches/something), just adding to the rampant speculation here for entertainment value. YMMV, IANAL, consult a doctor if it lasts more than four hours.**
 
  • Like
Reactions: nabsher, JerseyShoreMY, DudeByTheBeach and 11 others
#6,036
mehfisto said:
Did a little digging to figure out where the tesla site is running. Turns out it seems to be hosted on a front end IP address that is owned directly by them. The ASN for the IP that tesla.com resolves to matches tesla as well so this is advertised by them direct to the internet. It does traceroute through akamai, so there's some CDN in front of it for volumetric protection or WAF, but it really seems based on a dig/traceroute/whois/ASN lookup that this site is hosted from their premises, not in a cloud. It could be proxying traffic via load balancers/api gateways from web services in a cloud, but this could explain capacity issues, etc, like a bunch of data center appliances that are melting (WAFs/firewalls that are maxed out on flows/etc), or capacity in a blade chassis/hw cluster that normally can handle it. It wouldn't surprise me either if they were querying an ERP that manages the assembly line for the status of each order, and that ERP has an API query limit that is exhausted by all of us refreshing all day long. I'd put money on the fact that at least some of the outages are failed code pushes trying to figure out how to do a cache layer on that API, or something to offload that traffic. Doubt it's the same crew designing the car facing stuff. Although, I recall a reddit thread from a few years ago that talked about some basement duct tape/bubblegum infra that used to run things. I've seen other stuff saying they've moved to Kubernetes, etc for the car facing apps, but this account site could be a lot of tech debt that is now coming due for payment.

**I'm a cloud nerd who doesn't want to write documents for work right now, so I did this instead. Not making excuses for them (at least just put up a page that throttles requests/static caches/something), just adding to the rampant speculation here for entertainment value. YMMV, IANAL, consult a doctor if it lasts more than four hours.**
Click to expand...
I'd concur it's a problem to the backend, either middleware between ERP or directly to ERP. We used to do this to the mainframes (Aerospace, big Blue and Fujitsu still rule there along with Un*x big boxes) but I'd expect Tesla to be on something more modern than that lol.

Still, Front end to Back end middleware is obviously the problem since only deeper car details truly fail when I hit it. Akamai is probably just images and whatnot. Not sure if it'd be query limit or the load on whatever backend ERP they use (and I'm guessing they use one of the big Auto ERPs so maybe Epicore, Oracle or SAP -- they'd be absolutely insane to do home grown MRP/ERP/SFC) -- if they are using someone's cloud ERP it's probably pipe width between them and partner or front facing cloud eCommerce and partner.
 
#6,039
mehfisto said:
Did a little digging to figure out where the tesla site is running. Turns out it seems to be hosted on a front end IP address that is owned directly by them. The ASN for the IP that tesla.com resolves to matches tesla as well so this is advertised by them direct to the internet. It does traceroute through akamai, so there's some CDN in front of it for volumetric protection or WAF, but it really seems based on a dig/traceroute/whois/ASN lookup that this site is hosted from their premises, not in a cloud. It could be proxying traffic via load balancers/api gateways from web services in a cloud, but this could explain capacity issues, etc, like a bunch of data center appliances that are melting (WAFs/firewalls that are maxed out on flows/etc), or capacity in a blade chassis/hw cluster that normally can handle it. It wouldn't surprise me either if they were querying an ERP that manages the assembly line for the status of each order, and that ERP has an API query limit that is exhausted by all of us refreshing all day long. I'd put money on the fact that at least some of the outages are failed code pushes trying to figure out how to do a cache layer on that API, or something to offload that traffic. Doubt it's the same crew designing the car facing stuff. Although, I recall a reddit thread from a few years ago that talked about some basement duct tape/bubblegum infra that used to run things. I've seen other stuff saying they've moved to Kubernetes, etc for the car facing apps, but this account site could be a lot of tech debt that is now coming due for payment.

**I'm a cloud nerd who doesn't want to write documents for work right now, so I did this instead. Not making excuses for them (at least just put up a page that throttles requests/static caches/something), just adding to the rampant speculation here for entertainment value. YMMV, IANAL, consult a doctor if it lasts more than four hours.**
Click to expand...
They don’t use an ERP. They use their own built-in-house platform developed by their former CTO. It’s called WARP. I think.
 
You must log in or register to reply here.

Similar threads

C
Inventory Order Model Y, mentions it can be delivered to closer Tesla service center - no mention of cost and time
Replies
9
Views
474
Model Y: Ordering, Production, Delivery
CoolBuddy
C
H
FSD Transfer Question
Replies
18
Views
1K
Model 3: Ordering, Production, Delivery
ken_neth
ken_neth
S
Tesla delayed delivery and refuses to honor agreed to price?
Replies
16
Views
817
Model Y: Ordering, Production, Delivery
Leeclanual
Leeclanual
V
Manufacture Date & Delivery Date
Replies
8
Views
423
Model Y: Ordering, Production, Delivery
EVChris
E
Sherwood_John
Man, times have changed.
Replies
13
Views
1K
Model Y: Ordering, Production, Delivery
mikeskuro
M
Top
Back
Blog
New
Forum list
Marketplace
Menu