We announced LiteFS Cloud a few months ago as a way for SQLite users to keep their data safe with continuous backups and point-in-time restores. We originally released it in “preview” to let folks try it out and thank you to everyone who’s given us feedback!
We’re excited to announce that we’re moving LiteFS Cloud to be generally available!
Starting next month, LiteFS Cloud will cost $5/month for up to 10GB of data. If you need more data than that, no worries! It’s only $0.50 per GB after that. Data transfer rates outside the LiteFS Cloud region will be the same as our outbound transfer rates. It’s a great way to provide extra durability and peace of mind in case you accidentally forget a WHERE clause in your code.
If you’re building a platform on top of LiteFS and need much more space, get in touch! We have volume discounts for customers who need 500GB or more.
Very cool and congrats on reaching this milestone, but I’m disappointed in the $5 minimum in your pricing model. Would you consider perhaps some day dropping that minimum and simply price it at $0.50 per GB or even more granularly at $0.0005 per MB?
I agree. That pricing could be a stopper for new adopters and experimental projects. I have deleted one of my clusters because of it. I like @Alex21 suggestion.
PS: The cluster still appears in the dashboard after being deleted.
I can understand the appeal of pure usage-based pricing but it’s difficult to run a service and provide support without some kind of minimum. We may re-evaluate the pricing in the future though so it’s not set in stone.
@gcv Can you shoot me an email at ben@fly.io with the cluster name you’re trying to delete and I can take a look.
I can understand the appeal of pure usage-based pricing but it’s difficult to run a service and provide support without some kind of minimum. We may re-evaluate the pricing in the future though so it’s not set in stone.
Running litefs on 7 nodes over 6 months. 6 releases with no hiccups and one time a node crashed and recovered appropriately.
For reference, my DB is < 200MB and the backup is < 20MB compressed.
The damn thing just works. So for me, I’d have to really understand if I can truly make use of the PITR function meaningfully or if it is just extra complexity.
I’m glad to hear it’s running well! For the backups/PITR, it’s mostly a question of convenience and peace of mind. There’s other ways to back up LiteFS but LiteFS Cloud is probably the easiest.
We have some other features coming for LiteFS Cloud in the near future too. Are there features missing that you think would make it worth the cost?
For disaster recovery, DNS and backups need to be accessible outside of the hosting platform. We’re protecting against Fly getting DDOS’d, ransomware, or just accounting issues.
Having full backups that can be easily restored (customizable timeframe) without additional software is important. For example, I would like:
7 daily snapshots
12 weekly snapshots
24 monthly snapshots
The PITR can be hosted with fly, but other options are needed for DR / archival backups:
Customer provided bucket / storage and Fly managed backup rotation
Customer provided immutable bucket and Fly provides a script to manage backup rotation.
Right now, instead of PITR I take a snapshot every 6 hours which works due to my data size.
Having it available outside of Fly-controlled storage is a hard requirement for DR goals, so full control of storage medium for archival backups + PITR would entice me.
Your PostgreSQL is a shit show, even for non-production environments.
Is your SQLite going to be the same?
To elaborate: managed databases are outside of your core offering and your best shot with PostgreSQL is to partner with experts. You set expectations at the level of managed PostgreSQL service and failed to deliver.
Now, from an outsider’s perspective, you’re doing the same with SQLite.
Your PostgreSQL is a shit show, even for non-production environments.
I believe the correct word is terrifying. I’ve run database clusters for 15 years (mostly mysql) and this postgres + stolon + consul combination…I actually still don’t know how to recover it if something gets off the happy path, which is why I moved to sqlite/litefs.
Now, from an outsider’s perspective, you’re doing the same with SQLite.
I have a small database, but it has improved latency and hasn’t given me any headache. No opinion on litefs cloud as I haven’t used it. As far as I’m concerned, it’s infinitely better than the unmanaged postgres.
I respect that, and I want fly.io to succeed and be profitable and sustainable. But the core of your business already uses usage-based pricing. You can get a 1GB volume for $0.15/mo and run a 1GB RAM VM that scales to 0 and if it’s only receiving traffic for 5% of the time, that costs about $0.30/mo. It just seems like a big leap to then pay 10x that combined amount for a backup service (even a really cool backup service).
$5/mo might not seem like much, but it does add friction for students, hobbyists, tiny nonprofits, etc., especially in low income countries.
Also, there’s currently something like tens of millions of WordPress websites out there (as well as websites built with other software) on cheap and crappy shared hosting. Fly.io would be a much better option (more secure, more control over the software stack, ability to add geographical distribution) for running light traffic sites like that, especially if it included long-lived backups and PITR and was cheaper than what those site owners are currently paying.
Say I want to have a few instances of my app running (production, beta, and canary) that are run as separate fly apps, each with its own sqlite db. If I want to use LiteFS Cloud with them, how should I best structure them wrt clusters and databases?
Should each get its own cluster, with a server.db inside? Or alternatively, do I have a single cluster with production.db, canary.db, and beta.db inside? Does pricing change with what I do here? (i.e. have clusters are billed separately?)
Ideally, I don’t have to pay an additional $5/mo each for my beta/canary clusters that won’t see a lot of use, but are important for testing.
@benbjohnson I don’t know if this is the right place to ask this, my current month billing shows $5 for LiteFS Cloud, but I currently don’t have any clusters and I’m not even using LiteFS anymore on any of my apps.
I only used Cloud briefly when it was in preview, so it’s at least 2 months that my account doesn’t have any clusters active…
@tj1 Sorry for the late reply. No, you don’t need to do anything to upgrade from v0.4.0 to v0.5.x. There are some config sections that were added (backup & log) but no config updates were made.
I’ve run database clusters for 15 years (mostly mysql) and this postgres + stolon + consul combination…I actually still don’t know how to recover it if something gets off the happy path, which is why I moved to sqlite/litefs.
