Jump to content

Wednesdays and negative 9's uptime

Recommended Posts

I found out there is scheduled maintenance with Evernote on Wednesdays.  We live in a world where uptime was measured at the 99% level in year ~2000, then it was 99.9 which I thought was pretty amazing, then they came up with the 2 nines expression of 99.99, Three nines was not far off, and now, 99.9999% uptime is pretty much the standard.  Three nines is the bare minimum.


I don't know how long their downtime is, but if it is known and scheduled, I should not get an error in Evernote when it can't sync.  I should get a message that maintenance is happening, and sync has been disabled for x hours or something.  Especially if there is a timer that Evernote syncs on a schedule.  Why have the client pound on your systems when you know in advance that there is no point.


I feel like Evernote was a great desktop app and the cloud caught them off guard, they were not sure the cloud was going to be a thing ( I hate calling it the cloud, but just multiple client parity is what it is ) so they waited, a little too long, and had to throw something together. They used staff that did not have enough cloud infrastructure experience.  Most apps these days, from what I understand, dogfood their API.  Could Evernote build a desktop client that only used their own API?  Look at the twitter Desktop app, that is exactly what that is.  Twitter may have bought it, but before that, the developer had no special access.


I can't think of a single cloud based app that has scheduled downtime.  What in the world are they doing?  I can take a guess.  All data sits in a database.  Hopefully MySql, maybe Postgres, but probably neither.  They probably need a nosql mix with MySql, and to move it all over to AWS or similar and would save a ton of money too. Look what reddit saved by dropping their in-house system.  I have a feeling some of the servers are pretty old, but they do work.  However, this database they use, please don't tell me this is backed by something like filemaker, or something proprietary, or worse, they made something in house.  Whatever it is, it doesn't support replication or not easily.  It also probably can't do simultaneous reads and writes, so there is a ton of row locking going on.  That would explain some of the stalling I feel in sync.  


But they need a backup.  If Evernote went down, my guess, the restore would be as good as it was close to Wednesday.  If it were a Tuesday, we would lose all data up to last Wednesday.  They are performing backups.  And they will take longer the more data we put into their system.  A dump locks the entire database.  Ideally, you have a master, and tons of replication servers, you can toggle a replication server into master status, run your backups, and then toggle it back, it will start syncing up in the background.  There may be sharding of data all over the place as they outgrew the master, further complicating things.  I certainly hope that during this downtime they have tested a restore to make sure it actually works.


Oh, and their data feed from Level3, or their colo at Level3, that would be over with.  I can't support someone who is actively throttling Netflix in secret and refuses to accept their cache box from them, which is free, and also means doorstop delivery of data so no transport bandwidth fees.  The only reason you pass up on that deal is because you have ties to Comcast so you do what they tell you.  Claim Netflix is degrading performance, and throttle them down.  Now customers get annoyed, and by Comcasts logic, will move back to them.  We pay you for internet, get over it, sorry TV is dead, stop fighting it and figure out how to evolve into it.  Otherwise, you might as well just call the old CEO of Blockbuster and start commiserating right now.  Oh wait, I forgot, you don't care, your just a CEO who wants money, you aren't in it for the long haul, whereas, Netflix's owners care, because it is their baby.


Evernote, have you seen the amount of data Netflix posts about their infrastructure?  The blog posts that explain what software they use, how they have an app that randomly takes out nodes on their network to make sure they are resilient?  Do you know how much people appreciate that transparency?  Do you know how much people learn from it?  Have you seen their postmortems when things break and really go down?  It happens, it's not the end of the world.  By telling the users that x, and y happened, we can judge their competency.  When they fail due to incompetency, they admit it, and they explain how they have put in place policy or technology to mitigate it.  When they run into a really hard to troubleshoot problem, they postmortem that in case it will help others.  When Comcast and Level3 are in bed with each other and Netflix has proof, they let us know so we know who to blame.


I reported level3 issues to you, but I guess I did it wrong or something, because you want a log file from me, which tells me the tech did not understand, or like an insurance company that has a policy to deny first, and only resolve if the customer really gets angry and aggressive about moving it.  Is your policy no log file, no support, drop the ticket?  That assumes 100% of support issues are app related.  What if I am repotting a GUI glitch, what if I find a bug?  What if I am asking about a bug bounty and the policy?  What if I can log into anyones account?  I send in support, and you reply with, we need a log file.  Ok, cool, ill sell the bug off of on the dark net.  I would never do that, but it would piss many off enough to do exactly that.


It is 2015, once a week scheduled downtime is unacceptable.  Scheduled downtime is reserved for upgrades that improve the service, they are not a repeating event on a calendar. At 8 hours downtime once a week, I am not sure how long the downtime is, I never asked as I am sure it is proprietary data, and I don't want to sent a log file with personal data in it anyway, though i guess it doesn't matter, you can just look up my data at will.   That is 95% uptime!  Put that as a bullet point on the home page, I am certain it will attract a lot of new users.  I NEVER would have put all my data in this app had I known that and the current state of backup.  In this dat and age, most assume those two points are solved.  Now I am locked into this proprietary format, or I could try an applescript, or I could try your API, but after looking at it, it does not seem to be a REST based API, so it takes a bit to even play with it, rather than just opening a shell and pasting in a few curl commands and getting back some JSON.  If the API is anything like my experience with the app, for a one off exporter, it's just not worth the time.


Speaking of the log file, why do you need to store personal data in it?  I am pretty sure each message has a GUID to it, store that.  You don't need the subject.  If you do, add a debug mode into the prefs that turns on subject logging, explain why you need it, and allow the user to decide.  I can see no reason why you need it aside from perhaps localization issues, but you are logging to a plain text file, so your character support is probably limited anyway.


The new web based Evernote, how did that even get called a beta?  It is still alpha, it is not feature complete by any means.  It is not in parity with the app.  The notebooks don't even sort in the same order.  My inbox is at the bottom on the web, and top in Evernote.app.  Did anyone print once from the web?  Try it, print to a pdf so you don't waste the paper.


I left Things for Evernote because it has such a huge following.  I want to use this app, but I don't know if I should hold out because you guys say little.  Get a support system that can use email so we can email support@ and have a tech look at and reply to all.  I had a 60+ thread with Agile Software, and I have told everyone, last I remember, 8 people bought 1Password.  I would not recommend Evernote in it's current state.  Support is how word of mouth works in software these days to a large degree.  Some care so much they ask you to survey, some care more and ask you to type in your own words so you know they are reading them, especially since they reply if you have any issues.  


I should have read these forums first.  But Evernote rides on the wave of it's nice and simple past.  Eventually waves crash on the shore. Things is very tight lipped, and that was why I left, not because of anything in the app, their app actually is solid, so being tight lipped was not a show stopper.  But they weren't that tight lipped either.  


I really do want to keep using Evernote, I see potential, but lock in, scared of data corruption, adding a note via quick note to only wait seconds to launch the app to double check if it made it defeats the purpose of the quick note.  And yes, many a time a note has not made it.  That is why I asked where the quick note stages it's files, so I could perhaps watch that area rather than wait so long for the app to launch.  With each release, launches get slower.  I didn't even try OneNote, launched it, took 6 seconds, put it in the trash.  Just because we have fast CPU's doesn't mean we shouldn't develop as if we didn't. I don't see why older developers are having trouble getting work.  New developers never had to eek out every bit of performance.  Would any kid out of college know how to do a spell checker when the dictionary word file is larger than all the ram in the system?  Now we have so much ram we just stuff it in there, why?  Why not use the past efficient way? They don't know about it, they weren't born yet.  Prosily, I think every team should have an old grey beard as a project manager.


Stop reading here if you don't care about rants and pipe dreams :)


I will go out on a limb, if somehow a kickstarter could get me 300K, that would get me two developers, possibly a designer, though not sure I need one, it's an email client with read and write abilities.  100K per developer, and 100K left to build out the AWS redundant infrastructure. I will tell you right now, I would rip out a bunch of features, chat is gone, too many chat colab apps out there.  But I would add in SubEthaEdit style note sharing i.e.: more than one person typing on the app at the same time. The IOS app would either not exist, or be very minimal in the beginning.  The first thing I would do is build the API, get it out there, and hope someone else would make a nice iOS app like what happened with twitter and a few other companies.  The web and app store version would be identical.  It would auto backup on schedule.  It would read and write from one location, so if you don't want to trust our backend for sync, drop it in dropbox, one drive, etc.  At your peril of course, because I want to support meta data and not corrupt files.  And yes, Open note in x.app, so you can edit a note in your favorite editor, save in that editor, and it will save to your note. Other stuff, but outside that, it will be your basic panel of folders, a list of items in those folders, and a preview/edit window.  Full text search, a feature in databases that gives you great search with relevancy rankings.  Even Soundex support.  Data de-duplication to save on storage fees and sync times.  It won't be a photo storage app, it won't have a web app, it won't be an email client, or email storage app, you won't want to use it to manage your Garbage Pail Kids collection, but the API will allow anyone to make an add on if they desire.  It will use open formats and open source tools wherever possible, encrypted in a way even I can not read your data, and I will answer any question you want, hell, I will even open our accounting books to a github account.  Basically, I will make a core, features can be suggested, get 90% community approval, and of course, it's getting added, until the first time I hear the word bloat, at which point, no new features, and code review kicks in.  Apple's OS's are free, I suspect the app will only officially be supported on the current OS plus one behind it.   Apple moves fast, I am not getting stuck like others with Carbon, ATSUI type issues, resolution issues, etc.


Sorry, had to be said.


Link to comment


This topic is now archived and is closed to further replies.

  • Create New...