Percona Live Europe Amsterdam Day 1 notes

Percona Live Europe Amsterdam Day 1 was a bunch of fun, especially since I didn’t have to give a talk or anything since my tutorial was over on Day 0.

At lunch, I realised that there are a lot more fringe events happening around Percona Live… and if you’ve seen how people do “tech weeks”, maybe this is what the event ends up being – a show, plus plenty of focused satellite events. FOSDEM in the open source world totally gets this, and best of all, also lists fringe events (see example from 2019).

So, Thursday evening gets a few fringe events, a relatively short train ride away:

Anyway, what was Day 1 like? Keynotes started the morning, and I did make a Twitter thread. It is clear that there is a lot of talk amongst companies that make open source software, and companies in the ecosystem that somehow also derive great value from it. Some look at this as the great cloud vendors vs open source software vendors debate, but this isn’t necessarily always the case – we’ve seen this as Percona’s model too. And we’ve seen cloud companies contribute back (again, just like Percona). Guess this is a topic for a different post, because there are always two sides to this situation…

It is also clear that people want permissive open source licenses over anything source available. If you’re a CxO looking for software, it would be almost irresponsible to be using critical software that is just source available with a proprietary license. After all, what happens when the company decides to ask for more money? (Companies change ownership too, you know).

It is probably clear the best strategies are the “multi” (or hybrid) strategies. Multiple databases, multiple clouds, and going all in on open source to avoid vendor lock-in. Of course, don’t forget that open source software also can have “vendor lock-in” – always look at the health metrics of a project, vs. a product. We’re lucky in the MySQL ecosystem that we have not just the excellent work of Oracle, but also MariaDB Corporation / MariaDB Foundation and also Percona.

MySQL 8.0 adoption is taking off, with about 26% of the users on it. Those on MySQL 5.6 still seem to be on it, and there has been a decrease in 5.7 use to grow that 8.0 pie. It isn’t clear how these stats are generated (since there is no “phone home” functionality in MySQL; also the MariaDB Server variant doesn’t get as much usage as one would like), but maybe it is via download numbers?

Anyone paying any attention to MySQL 8 will know that they have switched to a “continuous delivery model”, also known as, you get new features in every point release. So the latest 8.0.18 gets EXPLAIN ANALYZE, and while we can’t try it yet (not released, and the documentation isn’t updated), I expect it will be fairly soon. I am eager to try this, because MariaDB Server has had ANALYZE since 10.1 (GA – Oct 2015). And it wasn’t long ago that MySQL received CHECK constraints support (8.0.16). Also the CLONE plugin in 8.0.17 warrants some checking/usage!

Besides all the hallway chats and meetings I did manage to get into a few sessions… Rakuten Intelligence talked about their usage of ProxySQL, and one thing was interesting with regard to their future plans slide – they do consider group replication but they wonder what would replace their custom HA software? But most importantly they wonder if it is stable and which companies have successfully deployed it, because they don’t want to be the first. Question from the floor about Galera Cluster came up, and they said they had one app that required XA support – looks like something to consider once Galera 4 is fully baked!

The PXC–8 talk was also chock full of information, delivered excellently, and something to try soon (it wasn’t quite available yesterday, but today I see a release announcement: Experimental Binary of Percona XtraDB Cluster 8.0).

I enjoyed the OpenCorporates use case at the end too. From the fact that for them, being on-premise would be cheaper than the cloud, how they use ProxySQL, Galera Cluster branch Percona XtraDB Cluster (PXC), and ZFS. ZFS is not the most common filesystem for MySQL deployments, so it was interesting to see what could be achieved.

Then there was the Booking.com party and boy, did they outdo themselves. We had a menu, multi-course meal with wine pairings, and a lot of good conversation. A night wouldn’t be complete without some Salmiakkikossu, and Monty sent some over for us to enjoy.

Food at the Hilton has been great too (something I would never really want to say, considering I’m not a fan of the hotel chain) – even the coffee breaks are well catered for. I think maybe this has been the best Percona Live in terms of catering, and I’ve been to a lot of them (maybe all…). I have to give much kudos to Bronwyn and Lorraine at Percona for the impeccable organisation. The WiFi works a charm as well. On towards Day 2!

ProxySQL Technology Day Ghent 2019

Just delivered a tutorial on MariaDB Server 10.4. Decided to take a closer look at the schedule for Percona Live Europe Amsterdam 2019 and one thing is clear: feels like there should also be a ProxySQL tutorial, largely because at mine, I noticed like 20% of the folk saying they use it.

Seems like there are 2 talks about it though, one about real world usage on Oct 1, and one about firewall usage with AWS, given by Marco Tusa on Oct 2.

Which led me to the ProxySQL Technology Day 2019 in Ghent, Belgium. October 3 2019. 2 hour train ride away from Amsterdam Schipol (the airport stop). It is easy to grab a ticket at Schipol Plaza, first class is about €20 more per way than second class, and a good spot to stay could be the Ibis Budget Dampoort (or the Marriott Ghent). Credit card payments accepted naturally, and I’m sure you can also do this online. Didn’t take me longer than five minutes to get all this settled.

So, the ProxySQL Technology Day is free, seems extremely focused and frankly is refreshing because you just learn about one thing! I feel like the MySQL world misses out on this tremendously as we lost the users conference… Interesting to see if this happens more in our ecosystem!

Moving images from an iPhone quickly

Wanted to move 60GB of photos from an iPhone to the Mac. It turns out that Dropbox does not save Live Photos, Google Photos does, but it is all in reduced size (unless you pay them).

Extraction via Image Capture was taking forever.

Had to make sure on the iPhone, in Settings, you select “Keep Originals” under Photos -> Transfer to Mac or PC.

Didn’t realise Image Capture also has this option, it is small, at the bottom, but it also exists.

Fast, HEIC images, with accompanying Live Photos (i.e. movie files).

Lyft and the bet on the millennial mindset

I’ve been following Lyft going public, and its clearly nuts. For an analysis of the S1, do read some Fred Wilson: The IPO Price and the S1.

My belief is that people are excited by tech IPOs, rather than the prospects of Lyft. When is it expected to be profitable? Do not use the Amazon argument, one is an Everything Store (a great read by Brad Stone), and the other is a provider of transport in limited markets.

Personal journey with Lyft: had not installed it in March 2015, still not in May 2016, tried it (i.e. downloaded it) July 2016 and deleted it a week later! There was at some stage a partnership where you could order a Lyft from within the Grab app, but that never worked (July 2016, March 2017).

What helped Lyft grow might have been the #DeleteUber campaign of 2017, though I still continued to use Uber. It was quite prevalent in San Francisco. Uber rates going up 10% in San Francisco still failed to push me to become a Lyft user.

However, a16z was betting on the “millennial mindset” when they poured money into Lyft, read their thesis/blog post. As I said in 2015:

“The bet on Lyft has always been that ridesharing wins” read the investment thesis. A real bet on “millenial mindset”

Did it pan out well? Sure, millennials on the Robinhood app traded it a lot upon IPO. San Francisco folk (maybe with some virtue signalling?) traded it 3x more than those in New York City. Fascinating piece in Axios: How millennial investors reacted to Lyft’s IPO. Maybe millennials treat Lyft (and tech IPOs?) like cryptoasset investing?

Anyway, FT has covered it best: But today, it seems Mr Market has woken up to realise Lyft is trading at 10x sales with negative 44.3 per cent ebitda margins and is operating in a market with no proof of profits. in Lyft is nuts, and it’s crashing.

The future will be very exciting, I expect more consolidation in this space (don’t all long tentacles go back to SoftBank?). If you have access to The Information, they’ve got an exclusive on Lyft’s Biggest IPO Winners. Just one highlight:

Sean Aggarwal, Lyft’s first outside investor, who put in $30,000 in 2007 and has a stake now worth $100 million. The biggest venture capital winner may be Floodgate Ventures, a seed investor which has seen a nearly 10,000% return on its roughly $1 million investment. Then there is Mayfield Fund, which invested a total of nearly $15 million in the first four venture rounds, giving it a stake likely to be worth about $600 million at the IPO.

State of the Mac early 2019

I landed in London sometime in January 2019, and made an appointment at the Regent Street Apple Store the same morning for that afternoon at 1pm. This is apparently a very rare thing, because Apple Store appointments are usually full up weeks in advance.

The reason was simple: my 2016 15” retina MacBook Pro was starting to wobble at the base and the battery indicator did say it was time for a service. This battery for what it is worth has only done under 200 cycles, so it did seem odd. The wobbling suggested that the battery possibly had expanded, and this was also affecting my use of of some keys — notably the left shift key, and on the right side the ? Key (it would many times skip giving me ? But instead give me /).

I arrived for my Apple Store appointment a little early and was told that I could only check in at 12:50pm. Fair enough, I would just wait it out at the store. The Genius Bar was packed full. I was finally told to take a seat upon checking in and a genius would be with me shortly.

I met a friendly genius, who told me that it is likely the battery expanded and they would have to change the whole top cover as this was all integrated, meaning I would also get a new keyboard. Then he said that I would need to typically leave it there for up to 7 days, but since I was travelling all around, they would try to do it within 24 hours. 

As luck would have it, they did not have the parts available in store, and then checked stock to see that Covent Garden had 2 units. Upon calling, they would take 2 weeks (yes, a whole 14 days) to fix it. This was unacceptable. The genius wrote up a report about how they were willing to do it at Regent Street and how this was odd that Covent Garden would not in a timely fashion; he also encouraged me to get this fixed ASAP as it was not a good idea to continue using said laptop. This is great news to hear when travelling, clearly.

So it is clear: Apple laptops are still not made for Enterprise use. Fast forward a couple of weeks later and I order a Mac Mini to my home in Malaysia. I figure it is a computer I have wanted to get for sometime and I would use a desktop for the one week or so that I would be in town.

I make a Time Machine backup of my MacBook Pro, and restore it to my 12” MacBook (2016) that I have not sold from a previous engagement. The restore takes many hours, and when it is finally done, the one thing that I consider quite important — Mail — is not fully migrated. All the local folders are not converting, and maybe this is a Mojave bug (I was coming from High Sierra). So I am thankful to start exporting Mail in the MBOX format. In addition, Chrome seemed to be missing all the plugins that I had installed, so this was further manual work.

This leads me to believe that when I do setup the Mac Mini and also when the MacBook Pro comes back from a service, I will set it up fresh (or hopefully, the MacBook Pro just needs an update, and the data remains intact; this is apparently true according to the genius — nowadays Apple does not even ask for your password any longer). After all, I do have iCloud sync turned on (so Keychain is synced), Dropbox seems to be doing more work than expected, and passwords are managed in 1Password. I’ll have to sync my Mail manually, which seems like some bandwidth will be used, but that seems fine. But maybe I won’t be needing all the applications that I have. And what about my home directory? (The Linux advantage is just moving your home directory; I am uncertain if this is true with macOS as I have always relied on Time Machine).

Today I visited the service centre. I’m told that it would be there possibly for the 7 day period, so I may be able to get it back next Saturday. If this were my only Mac, I would be out of commission for a whole work week. If I had purchased a Dell or a Lenovo, I would have all this fixed by the next business day. Alas, I am still “locked into” macOS.

Initial Mojave thoughts? Kind of silly that when I fire up Terminal I can’t even do a ls in ~/Library/Mail unless I give Terminal all disk access. What a mess, all this lockdown is, if you ask me. 

I’ll hope to have a positive update later this week. The Mac Mini I ordered only comes with 8GB of RAM, since the Apple uplift for 2 16GB sticks (32GB of RAM) would have added RM2,640. I asked the Apple Service Centre how much it would cost, and they too said about RM2,000 since they have to order it from Apple (at first they said it could not even be upgraded!). I’ll go the after-market route, where each stick is only RM490, so getting it all for RM980 seems like a better choice. I’m going to guess that the Mac Mini 2018 model isn’t selling all too well in Malaysia, judging by many saying they haven’t upgraded it yet…

Tech teams and hiring right

“In the end we couldn’t find a tech team we could trust and that could produce what we asked of them.”

How these words ring so true. This is from a Medium post by John Biggs, now working at The Block.

It is amazing to see what happened, as they hired two tech teams:

“Our first tech team worked mightily on product that worked but was not scalable. Our second tech team was self-interested and the tech did not work.”

They raised $150k, had 10 employees, two development teams, and over 8,000 beta users, but were unable to deliver functioning technology.

I have no idea if these were outsourced teams, or hired teams (I’ll go with outsourced), but this is a real problem facing many companies, not just those involved in the cryptocurrency space. Sometimes it also has to do with the irrational exuberance of the leadership team, directing the tech teams. This is where experience makes sense and pays off…

It makes me think a lot about the offering of the fractional CTO (basically a fractional CxO in general). There is a lot to do. A lot of patterns to recognise. A lot of project management. And a lot of understanding on what can and should be delivered in an appropriate timeframe. Something a really technical co-founder or lead will get right, but something the average founder will miss.


i