text stringlengths 0 1.49k |
|---|
**Brian Ketelsen:** Well, that means that you should feel awesome when somebody has a catastrophic failure and the backups work and they restore their data to a new system. So the corollary to that is that you should get to feel awesome pretty often, because people are restoring backups. |
**Alexander Neumann:** Exactly. I do, because people are creating backups, and this is so much better than not having backups at all. It must be really easy to do backups, because otherwise nobody does it. |
When I started with Restic, I had a look around and tried different backup programs, and there was this old approach to doing backups where you have to decide "Do I do a full backup or an incremental backup?" This was mental workload that -- as a user, I would just like to create a backup; I'm not interested in making ... |
**Ashley McNamara:** Oh, it's so annoying...! It's the reason we don't do backups. You're right. |
**Erik St. Martin:** I guess that was one of the reasons I love Time Machine. The first time I got a Mac with that on there... I only had to tell it where to store the backup data, and from there it kind of just did its thing. |
**Ashley McNamara:** You guys, I lost my Synology in the divorce... Sad times for me. \[laughter\] And my time machine is like "You haven't backed up in like 600 days", or something like that. I'm like "Yeah, I know..." Way to remind me, guys... |
**Brian Ketelsen:** It's time to get a new NAS. |
**Ashley McNamara:** I know! I can do better this time... On both accounts, probably. |
**Alexander Neumann:** Yeah, so after having a bit of research done for other backup programs I started a list of open source backup programs that work on Linux, and I keep discovering new ones... The list already has like 80-90 entries, something like that. It's on [github.com/restic/others](https://github.com/restic/... |
This is written in Python mostly, and it also had a C component and uses OpenSSL for the crypto. This is also very nice, but sometimes the workflow is not so great, because when you create a backup and there's this Borg Create command, then you need to give it a name. And every time I'm trying to use it, I'm thinking a... |
**Brian Ketelsen:** I like the workflow of Restic. It seems like you thought through nicely the idea of having multiple different repositories, and being able to backup to different repositories makes my life really easy. I like that a lot. |
As far as user interface and user experience goes, Restic is good. |
**Alexander Neumann:** Okay, that's very nice to hear, because this is really the focus of Restic. |
**Brian Ketelsen:** It is hard to do with a backup software -- I mean, it's backup software, it's not an easy thing. So it's nice that Restic is... |
**Alexander Neumann:** Yeah, indeed. |
**Brian Ketelsen:** But you use [Viper](https://github.com/spf13/viper) for the command line interface now and that helps a lot too, Viper is about as good as it gets on the command line side of things in Go. |
**Alexander Neumann:** Yeah, I thought of writing my own framework and then found Viper, and it does all that I need it to do right now. What's missing from Restic at the moment -- no, I think [Cobra](https://github.com/spf13/cobra) is the CLI framework and Viper is the configuration framework. |
**Brian Ketelsen:** Oh, you're right. |
**Alexander Neumann:** This Viper thing is a bit scary at the moment -- I don't know how it effectively works, and in terms of configuration files, Restic doesn't have a configuration file yet, because I haven't found a way to do this nicely. I prefer that people write their own shell scripts to run Restic, instead of ... |
**Brian Ketelsen:** That makes good sense. |
**Ashley McNamara:** Yeah, it's good advice. |
**Erik St. Martin:** What would happen if somebody changed their key in the middle of a backup? Just thinking of things that could go wrong with config files... |
**Alexander Neumann:** Yeah, this is an interesting question. What I was talking about was like a local configuration file where you say like "Oh, this is my backend, this is my exclude list for this directory" and so on. When you change the key, in Restic there is only just two keys for one repository. Whenever you in... |
It's especially important - the password is independent of the key. You can change your password and still access the old data that has been saved in the repo weeks before. So there is no key that you can change. |
There was a GitHub issue of somebody who said "Oh, we need to have cipher agility and be able to re-encrypt the complete repository and being able to use another algorithm instead of IAS, and this was something that I would not like to do. I think too many knobs for users and too many different code paths and algorithm... |
For Restic, I would always say that it shows a same default and make things configurable for users that need to be, but if in doubt, there won't be a knob for it. |
**Erik St. Martin:** So I think that we've probably overshot our sponsored break a little, so let's go ahead and take that real quick. Our sponsor for today is Toptal. |
**Break:** \[37:33\] |
**Erik St. Martin:** And we are back, talking to Alexander Neumann from Restic. Before we went to break we were talking about some of the feature requests and things that you've had... What's next for Restic? What are some things that you do want to implement? Where do you see this going? |
**Brian Ketelsen:** Yeah, so I did the 0.6.0 release today, the release candidate, so the development for that is done. The next projects are getting more backends into Restic. There's a Swift backend; this is some kind of object storage thing that you can also rent from OVH, for example. There's the Backblaze backend ... |
For example, I would like to support compression, because at the moment, when the file is read and spit into blobs, and these blobs are encrypted and saved the way they are into the repository... So having compression may be very efficient, because sometimes blobs can be compressed very efficiently. This is not support... |
Other parts of the project - we have a huge list of things that need to be reworked. It's the same, I think, with every non-trivial program. One of the things is that some operations for Restic are not so fast as they could be, and mainly this is a problem that we don't cache any data locally. Restic does not have any ... |
Some operations, like the prune operation - which goes through the list of all the blobs and looks for blobs that are not in use anymore but are not referenced anymore, and this is very time-consuming because it traverses all the tree structures in JSON and requests all the trees from the repository... So this is somet... |
I have many ideas which can also be made better, but this is the case with all the open source programs out there, I think. |
**Erik St. Martin:** So here's an interesting question, because there's some uniqueness to the way this application works... So how are upgrades performed? Is it that you've kind of pre-built into the storage layer some information for the tool to be able to kind of work backwards-compatible with old ways that the data... |
**Alexander Neumann:** Yeah, there will be a migrate command, which is able to convert a repository to a newer version - maybe also to an older version, I haven't decided that yet. |
At the moment - and I think that this is one of the most astounding things, at least for me - almost the first version of the repository that I released two-and-a-half years ago is still working today, so we haven't really changed much over there. The repository always has a configuration file which is also encrypted, ... |
At the moment Restic tries to access the repository, it downloads the configuration file, decrypts it and checks whether the configuration file has a version number that is compatible to the currently-run binary. So when we do the next version of the repository format, then there will be a 2 instead of a 1 in the versi... |
I'd like it to be as compatible as possible, because when people start using Restic, then they depend on us - mostly on me, in this case - but they are able to restore their backups in like ten years or so. For Go it's really great, because you can always say like "Okay, if you need to handle version 1 repositories and... |
This is really great, because you don't need any configuration, any libraries, any runtime, anything. It's all built into the binary. For the most part, we'd like to be as compatible as possible. |
**Erik St. Martin:** It's great that you at least really thought of that with the version stuff. |
**Alexander Neumann:** Yeah, this came out of many discussions with my colleagues at work, because they're also quick thinkers and we are also very interested in having a working backup program, so this is something that at work we sometimes saw failures. For example with TLS, version detection has been a problem, and ... |
**Erik St. Martin:** That's awesome. So I think we probably have about 15-20 minutes left, and I know the past couple of weeks we skipped over all the projects and news, and even \#FreeSoftwareFriday last week. |
**Brian Ketelsen:** Wow, that's against the law. |
**Erik St. Martin:** So do you guys wanna jump into anything there? I know we have kind of like a lot of stuff that's probably piled up. |
**Ashley McNamara:** I didn't contribute anything to the news, but I can talk to [Kelsey Hightower's](https://twitter.com/kelseyhightower) DevOps Days speech. |
**Erik St. Martin:** That's right, you were there. |
**Ashley McNamara:** I was there, in person, crying like a baby. |
**Brian Ketelsen:** Aw... |
**Alexander Neumann:** I skipped over the [video](https://youtu.be/36S7N7OZSTI) a few minutes ago and it seems to be that he was very agitated. I haven't seen it yet. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.