<antirez>

Twilio just released a post mortem about an incident that caused issues with the billing system:

http://www.twilio.com/blog/2013/07/billing-incident-post-mortem.html

The problem was about a Redis server, since Twilio is using Redis to store the in-flight account balances, in a master-slaves setup, with multiple slaves in different data centers for obvious availability and data safety concerns.

This is a short analysis of the incident, what Twilio can do and what Redis can do to avoid this kind of issues.

Yesterday night I returned back home after a short trip in San Francisco. Before memory fades out and while my feelings are crisp enough, I'm writing a short report of the trip. The point of view is that of a south European programmer exposed for a few days to what is probably the most active information technology ecosystem and economy of the world.

Reaching San Francisco
===

If you want to reach San Francisco from Sicily, there are no direct flights helping you. My flight was a Lufthansa flight from Catania to Munich, and finally from Munich to San Francisco. This is a total of 15 hours flight, plus the stop in Munich waiting for the second flight.

Redis uses streamed asynchronous replication, that's one of the simplest forms of replication you can imagine: a continuos stream of writes is sent to the slaves, without waiting for the slaves to process the writes in any way before replying to the client.

I always gave that almost for granted, as I always assumed Redis was not a good match for synchronous replication, that has an higher latency. However recently I tried to fix another issue with Redis replication, that is, timeouts are all up to the slave.

Terah is a planet far away, where networks never split. They have a single issue with their computer networks, from time to time, single hosts break in a way or the other. Sometimes is a broken power supply, other times a crashed disk, or a software issue completely blocking the system.

The inhabitants of this strange planet use two database systems. One is imported from planet Earth via the Galactic Exchange Program, and is called EDB. The other is produced by engineers from Terah, and is called TDB. The databases are functionally equivalent, but they have different semantics when a network partition happens. While the database from Earth stops accepting writes as long as it is not connected with the majority of the other database nodes, the database from Terah works as long as the majority of the clients can reach at least a database node (incidentally, the author of this story released a similar software project called Sentinel, but this is just a coincidence).

In a great series of articles Kyle Kingsbury, aka @aphyr on Twitter, attacked a number of data stores:

[1] http://aphyr.com/tags/jepsen

Postgress, Redis Sentinel, MongoDB, and Riak are audited to find what happens during network partitions and how these systems can provide the claimed guarantees.

Redis is attacked here: http://aphyr.com/posts/283-call-me-maybe-redis

I said that Kyle "attacked" the systems on purpose, as I see a parallel with the world of computer security here, it is really a good idea to move this paradigm to the database world, to show failure modes of systems against the claims of vendors. Similarly to what happens in the security world the vendor may take the right steps to fix the system when possible, or simply the user base will be able to recognize that under certain circumstances something bad is going to happen with your data.

Lately I'm trying to push forward Redis 2.8 enough to reach the feature freeze and release it as a stable release as soon as possible.
Redis 2.8 will not contain Redis Cluster, and its implementation of Redis Sentinel is the same as 2.6 and unstable branches, (Sentinel is taken mostly in sync in all the branches being fundamentally a different project using Redis just as framework).

However there are many new interesting features in Redis 2.8 that are back ported from the unstable branch. Basically 2.8 it's our usual "in the middle" release, like 2.4 was: waiting for Redis 3.0 that will feature Redis Cluster (we have great progresses about it! See https://vimeo.com/63672368), we'll have a 2.8 release with everything that is ready to be released into the unstable branch. The goal is of course to put more things in the hands of users ASAP.

Questo post ha lo scopo di presentare alla comunita' italiana interessata ai temi della programmazione e delle startup un progetto nato attorno ad un paio di birre: "Hacking Italia", che trovate all'indirizzo http://hackingitalia.com

Hacking Italia e' un sito di "social news", molto simile ad Hacker News, il celebre collettore di news per hacker di YCombinator. A che serve un sito italiano, e in italiano se c'e' gia' molto di piu' e di meglio nel panorama internazionale? A mettere assieme una massa critica di persone "giuste" in Italia.

Hello! As promised today I did some SSD testing.

The setup: a Linux box with 24 GB of RAM, with two disks.

A) A spinning disk.
b) An SSD (Intel 320 series).

The idea is, what happens if I set the SSD disk partition as a swap partition and fill Redis with a dataset larger than RAM?
It is a lot of time I want to do this test, especially now that Redis focus is only on RAM and I abandoned the idea of targeting disk for a number of reasons.

I already guessed that the SSD swap setup would perform in a bad way, but I was not expecting it was *so bad*.

One thing, more than everything else, keeps me focused while programming: never interrupt the flow.

If you ever wrote some complex piece of code you know what happens after some time: your mental model of the software starts to be very complex with different ideas nested inside other ideas, like the structure of your program is, after all.

So while you are writing this piece of code, you realize that because of it you need to do that other change. Something like "I'm freeing this object here, but it's connected to this two other objects and I need to do this and that in order to ensure consistent state".

After the "sexism gate" I started to use my Twitter account only for private stuff in order to protect the image of Redis and/from my freedom to say whatever I want. It did not worked actually since the reality is that people continue to address you with at-messages about Redis stuff.

But the good outcome is that now I created a @redisfeed account that I use in order to provide a stream of information to Redis users that are not interested in my personal tweets  not related to Redis. Anyway when I say some important thing regarding Redis with my personal account, I just retweet in the other side, so this is a good setup.

This is a very busy moment for Redis because the new year started in a very interesting way:

1) I finished the Partial Resynchronization patch (aka PSYNC) and merged it into the unstable and 2.8 branch. You can read more about it here: http://antirez.com/news/47
2) We finally have keyspace changes notifications: http://redis.io/topics/notifications

Everything is already merged into our development branches, so the deal is closed, and Redis 2.8 will include both the features.

I'm especially super excited about PSYNC, as this is a not-feature, simply you don't have to deal with it, the only change is that slaves work A LOT better. I love adding stuff that is transparent for users, just making the system better and more robust.

For a decade and half I contributed to open source regularly, and still it is relatively rare that I stop to think a bit more about what this means for me. Probably it is just because I like to write code, so this is how I use my time: writing code instead of thinking about what this means… however lately I'm starting to have a few recurring ideas about open source, its relationship with the IT industry, and my interpretation of what OSS is, for me, as a developer.

First of all, open source for me is not a way to contribute to the free software movement, but to contribute to humanity. This means a lot of things, for instance I don't care about what people do with my code, nor if they'll release back their modifications. I simply want people to use my code in one way or the other.

Dear Redis users, in the final part of 2012 I repeated many time that the focus, for 2013, is all about Redis Cluster and Redis Sentinel.

This is exactly what I'm going to do from the point of view of the big picture, however there are many smaller features that make a big difference from the point of view of the Redis user day to day operations. Such features can't be ignored as well. They are less shiny in a feature list, and they are not good to generate buzz and interest in new users, and sometimes boring to code, but they are very important from a practical point of view.

# Software defined radio is cool

About one week ago I received my RTLSDR dongle, entering the already copious crew of software defined radio enthusiasts.

It's really a lot of fun, for instance from my home that is at about 10 km from the Catania Airport I can listen the tower talking with the aircrafts in the 118.700 Mhz frequency with AM modulation, however because of lack of time I was not able to explore this further until the past Sunday.

My Sunday goal was to use the RTLSDR to see if I was able to capture some ADS-B message from the aircrafts lading or leaving from the airport. Basically ADS-B is a security device that is installed in most aircrafts that is used for collision avoidance and other stuff like this. Every aircraft broadcasts informations about heading, speed, altitude and so forth.

Currently I'm working on Redis partial resynchronization of slaves as I wrote in previous blog posts.

The idea is that we have a backlog of the replication stream, up to the specified amount of bytes (this will be in the order of a few megabytes by default).

If a slave lost the connection, it connects again, see if the master RUNID is the same, and asks to continue from a given offset. If this is possible, we continue, nothing is lost, and a full resynchronization is not needed. Otherwise if the offset is about data we no longer have in the backlog, we full resync.

While a big number of users use large farms of Redis nodes, from the point of view of the project itself currently Redis is a mostly single-instance business.

I've big plans about going distributed with the project, to the extent that I'm no longer evaluating any threaded version of Redis: for me from the point of view of Redis a core is like a computer, so that scaling multi core or on a cluster of computers is the same conceptually. Multiple instances is a share-nothing architecture. Everything makes sense AS LONG AS we have a *credible way* to shard :-)

Premise: a small rant about software reliability.
===

I'm very serious about software reliability, and this is not just a good thing.
It is good in a sense, as I tend to work to ensure that the software I release is solid. At the same time I think I take this issue a bit too personally: I get upset if I receive a crash report that I can't investigate further for some reason, or that looks like almost impossible to me, or with an unhelpful stack trace.

Guess what? This is a bad attitude because to deliver bugs free software is simply impossible. We are used to think in terms of labels: "stable", "production ready", "beta quality". I think that these labels are actually pretty misleading if not put in the right perspective.

This is one of this trivial changes in Redis that can make a big difference for users. Basically in the unstable branch I added some code that has the following effect, when running Redis on Linux systems:

[32741] 19 Nov 12:00:55.019 * Background saving started by pid 391
[391] 19 Nov 12:01:00.663 * DB saved on disk
[391] 19 Nov 12:01:00.673 * RDB: 462 MB of memory used by copy-on-write

As you can see now the amount of additional memory used by the saving child is reported (it is also reported for AOF rewrite operations).

Memory errors in computers are so common that you can register domain names similar to very famous domain names, but altered in one bit, and get a number of requests per hour:

http://dinaburg.org/bitsquatting.html

On Twitter:

@War3zRub1 "Hahaha it's silly how people use Redis when they need a reverse proxy"
@C4ntc0de "ZOMG! Use a real message queue, Redis is not a queue!"
@L4m3tr00l "My face when Redis BLABLABLA..."

Meanwhile *at* Twitter:

OP1: "Hey guys, there is a spike in the number of lame messages today, load is increasing..."
OP2: "Yep noticed, it's the usual troll fiesta trowing shit at Redis, 59482 messages per second right now."
OP1: "Ok, no prob, let's spawn two additional Redis nodes to serve their timelines as smooth as usually".

This post by Peter Bailis is a must read. "Safety and Liveness: Eventual consistency is not safe" [1].

[1] http://www.bailis.org/blog/safety-and-liveness-eventual-consistency-is-not-safe/

An extreme TL;DR of this is.

1) In an eventually consistent system, when all the nodes will agree again after a partition?
2) In an eventually consistent system, HOW the nodes will agree about inconsistencies?
3) In immediately consistent systems, when I'm no longer able to write? When I'm no longer able to read?

There is a new DB option out there, I know it took a long time to be developed. While I don't know very well how it works I hope it will be an interesting player in the database landscape.

My initial feeling is that it will compete closely with Riak and MongoDB (the system seems more similar to MongoDB itself, but if it can scale well multi-nodes people that don't need high write availability may pick an immediate consistent database such as RethinkDB instead of Riak for certain applications).

While I consider the Amazon Dynamo design, and its original paper, one of the most interesting things produced in the field of databases in recent times, in the Redis world eventual consistency was never particularly discussed.

Redis Cluster for instance is a system biased towards consistency than availability. Redis Sentinel itself is an HA solution with the dogma of consistency and master slave setups.

This bias for consistent systems over more available but eventual consistent systems has some good reasons indeed, that I can probably reduce to three main points:

You can follow the commits in the next days in the "psync" branch at github:

https://github.com/antirez/redis/commits/psync

From an HN comment[1]

"(Geek note: In the late nineties I worked briefly with a D&D fanatic ops team lead. He threw a D100 when he came in every morning. Anything >90 he picked a random machine to failover 'politely'. If he threw a 100 he went to the machine room and switched something off or unplugged something. A human chaos monkey)."

[1] http://news.ycombinator.com/item?id=4736220

I'm pretty surprised no one tried to write a wrapper for redis-rb or other clients implementing a Dynamo-style system on top of Redis primitives.

Basically something like that:

1) You have a list of N Redis nodes.
2) On write, use consistent hashing and write the same thing to M nodes (M configurable).
3) On reads, read from M nodes and pick the most common reply to return to the client. For all the non-matching replies, use DUMP / RESTORE in Redis 2.6 to update the value of nodes that are in the minority.

I'm proud to be mentioned in this well-thought and non-bigot post: http://www.nerdess.net/waffling/why-it-awesome-be-girl-tech/

Also in perfect accordance with the hacking culture, the post is sort of an HOWTO for girls that want to be involved in IT.

In this busy days I had the idea to focus on a single, non-huge, self contained project for some time, that could allow me to work focused as much as hours as possible, and could provide a significant advantage to the Redis community.

It turns out, the best bet was partial replication resync. An always wanted feature that consists in the ability to a slave to resynchronize to a master without the need of a full resync (and RDB dump creation on the master side) if the replication link was interrupted for a short time, because of a timeout, or a network issue, or similar transient issue.

I love Github issues, it is one of the awesome things at Github IMHO: as simple as possible but actually under the hood pretty full featured.

However one of the things I love more is labels. It is a truly powerful thing to organize issues in a project-specific way. Unfortunately if an issue is a pull request, no labels can be attached. I wonder why.

Also I would love the ability to merge against multiple branches instead of the taget one, directly from the web UI.

I assume you already read the AWS report[1] about recent troubles. I think it is a very good argument you could use at work against design complexity and in favor of designing stuff that are at a complexity level where analysis of failure modes and prevention is actually possible.

[1] https://aws.amazon.com/message/680342/

From a comment on Hacker News:

(link: http://news.ycombinator.com/item?id=4705387)

--- quoted comment ---
Full disclosure: I work for an AWS competitor.
While none of the specific AWS systemic failures may themselves be foreseeable, it is not true that issues of this nature cannot be anticipated: the architecture of their system (and in particular, their insistence on network storage for local data) allows for cascading failure modes in which single failures blossom to systemic ones. AWS is not the only entity to have made this mistake with respect to network storage in the cloud; I, too, was fooled.[1]

Achievement unlocked: releasing a Redis version the same day your daughter was born ;-)

But that was a bad issue as there was a bug preventing compilation on pretty old Linux systems that are still pretty widespread (RHLE5 & similar).

Redis 2.6.1 fixes just that issue and is available as usually at http://redis.io as a tar.gz or at github/antirez/redis as a "2.6.1" tag.

I really trust both in the usefulness of Redis bit operations and the fact that our community in the future should have documentation about Redis Patterns. So an article from CopperEgg where a bit operations pattern is described is good for sure :)

http://copperegg.com/redis-bit-operations-use-case-at-copperegg/

25 October 2012 01:06, she is 3350 grams of a funny little thing :-)

It's a more quite time now. Redis 2.6 released, the sexism issue almost forgotten. Time to relax, be wise, and focus on work. Right, but, that's not me. I've a few more things to say about what happened, and to reply to the many people that asked me why I felt "obligated" to stop using my Twitter account as before, with a mix of work, thoughts on technology, and personal stuff.

I can change idea easily if it is the case, but this time it was not the case. As much as people that criticised me for my blog post may think that I've a problem, I also think they have huge limits. Oh well, different opinions, I don't like you, you don't like me, I don't freaking care after all. I don't think on the same line as most people alive if that's the matter.

h2s writes about Linus:

"I love this guy's balanced approach to steering the kernel. Somebody asked whether a bunch of security-related patches would be getting into Linus' tree, and his response was great.
Basically, he spent a few minutes explaining how security people tend to think that problems are either security problems or not worth thinking about. They see things in black and white and only care about increasing security at any cost. He said performance fanatics can be the same in their approach to improving performance, and he tries not to treat security or performance patches as being too massively different from any other types of patches such as ones for correctness.

I don't like people that are using recent EC2 problems to get an easy advantage / marketing. Stuff go down and cloud services are not magical, it is better to adjust the expectations.

But there are other reasons why people IMHO should consider going bare metal.

* EC2 (and similar services) are extremely costly. With 100 euros per month you can rent a beast of a dedicated server with 64 GB of RAM and fast RAID disks.

* As you can see you are not down-time safe, and to be down together with a zillion of other sites may be a good excuse with your boss maybe, but does not change your uptime percentage, so it's a poor shield.

Redis 2.6 is finally out and I think that now that we reached this point we'll start to see how the advantages of a release that was already exploited in production by few, will turn into a big advantage for the rest of the community.

Scripting, bitops, and all the big features are good additions but my feeling is that Redis 2.6 is especially significative as a step forward in the maturity of the Redis implementation. This does not mean that's bug free, it's new code and we'll likely discover bugs especially in the early days as with every new release that starts to be adopted more and more.

And there is Pieter Noordhuis on stage right now!

http://redisconf.com/video/

You are there in the morning with your coffee in front of you, scanning pull requests and bug reports, then you see a conversation around a commit among a few guys that modified the code to make it better, than there is another one suggesting to improve it in another way. You click in the account names and you see this people with their transparent eyes and your trust in humanity is restored.

Takeaways:

1) Making videos is in some way harder than doing a talk live.
2) Screen Flow is awesome, but could be improved with more video editing capabilities, apparently you can't "cut" the video.
3) The problem is to upload files when they are big and you have normal ADSL connection :-)

But it feels good to be able to send the video talks a few days in advance, so the conf organizers will be able to perform editing, filter audio if needed, whatever.

of the final recording of the videos I'll send to the Redis Conf. That was hard! The timing of the conf was not excellent for my attending, but producing the video was also less trivial than I thought, but finally I've the slides, an idea about what to say, and the ScreenFlow skills ;) Maybe after this experience I'll produce some video tutorial of Redis new features as I introduce it, in order to accelerate the adoption of new things in our community.

Now back to work...

Good article by Josiah Carlson on Redis memory usage estimation, including copy-on-write worst case during BGSAVE:

https://groups.google.com/d/msg/redis-db/02oq_DNZA3s/l_uEwDT3d4sJ

On twitter I read a few concerns about inability to read what I think about tech non-redis topics. First of all, thanks to everybody interested in my thoughts :) Second, this blog is exactly the place where I'll post everything like that.

So:

@redisfeed -> Redis news, mostly low traffic, high signal.
@antirez -> Will be converted into my personal account, mostly italian language, non work related.
@zeritna -> Will be simply dismissed.
This blog -> Everything about day by day Redis development, personal opinion about sexism, sky driving, shit eating and japanese food.

I'm going to be a bit away from Redis code and this blog as I need to freaking focus on finishing the video talks for the Redis Conference that is very near at this point...

I'm also tuning and testing the final bits into 2.6 to make sure to release it ASAP :-)

See you in a couple of days.

p.s. also my wife feels contraction since a few hours, so maybe Greta is going to birth in a few... (!!!)

The commit message says it all:

    Fix MULTI / EXEC rendering in MONITOR output.
    
    Before of this commit it used to be like this:
    
    MULTI
    EXEC
    ... actual commands of the transaction ...
    
    Because after all that is the natural order of things. Transaction
    commands are queued and executed *only after* EXEC is called.
    
    However this makes debugging with MONITOR a mess, so the code was
    modified to provide a coherent output.
    
    What happens is that MULTI is rendered in the MONITOR output as far as

These are the real victims of flamewars:

"After the recent affair with @antirez's blogpost, I'm seriously considering forgoing further usage of #Redis in all my projects." (@nathell)

Obvious considerations:

1) Somebody will be happy about this as there was definitely a force trying to resonate as much as possible what was happening in the worst way.
2) The guy should read my blog post seriously.

But this is how it goes, you write a blog post that is a point of view about how to handle sexism (focusing on its effects, not the cause that anyway at the most subtle level, for instance a wrong promotion, can simply be denied by the author of the sexism), and the net result thanks to a number of people in part evil, in part simply bigot as hell, is that I'm sexist.

Hey, since I'm no longer active on Twitter, that was a channel where from time to time I replied to requests from users, I'm now making sure I reply to a few questions on Stack Overflow every day, lurking on the "Redis" tag:

http://stackoverflow.com/questions/tagged/redis

I can't ensure you that I'll reply to your question, but I can ensure you I'll be there often to reply to a few questions a day, and together we can create a Redis ecosystem in one of the best places on the internet where to give or receive help, that is Stack Overflow.

Now it's like a mix between a twitter timeline and a blog. I fixed the RSS feed, but still could be generated better than that. Well the point is that as far as I've the min needed to improve it in the future will be easy and fun.

p.s. yes the layout and fonts are a bit of a mess, but it's not going to be too hard to fix it. For now I focused more on what it should display.

Warning: I forgot to increment the js version counter so it takes a few hard reloads of the page to get the right CSS.

The new Twitter account @redisfeed will be used to provide informations about Redis new releases, critical bugs, and everything else is important for people that are using or plan to use Redis. Please follow us! http://twitter.com/redisfeed

Hi visitor! This blog was conceived for low traffic blogging. Now that I plan to don't use my Twitter accounts the old blog engine was not good enough.

The simplest thing to do was to take Lamer News and create a quick modified version that could be used as a blog engine... that's the result, for now. I hope to evolve it, but the point here is, I can write both long posts and very small ones that can be read directly from the home page.

It is also possible to write just short titles linking to external site, that is a feature I plan to use as well to link to interesting Google Groups comments and stuff like that.