-18

The site suddenly went into read-only mode today at ~13:25Z came without warning. The banner that goes along with read-only mode says:

This site is currently in read-only mode; we’ll return with full functionality soon. Follow @StackStatus or visit our status blog for more info.

Except, by 13:43Z, there had been no tweets from @StackStatus since November, and no posts on the status blog in three months. (Yes, I know it says "2 months ago". It's lying. The last post was written on 8th September.) So there was no explanation, either.

By now even the read-only site was gone, only a single line of default-font text reading:

Stack Exchange is currently offline, we'll be back shortly!

When's "shortly"?

Now, I appreciate that the team was likely busily trying to fix the fault — a tweet eventually surfaced at around 13:45Z claiming "database issues" — but could you perhaps do a better job of keeping us informed about maintenance, faults and the like?

At the very least, if there are no recent tweets or blog updates, the read-only banner shouldn't send us to Twitter or the blog for more information.

Improve this question
18
  • 6
    I wonder why people are downvoting this... it seems like a reasonable request.
    – David Z
    Dec 5, 2015 at 14:14
  • 9
    I'm not sure what you're asking here. Do you expect the unicorns to travel forward in time so they can tweet just before the freaking cluster blows up?
    – rene
    Dec 5, 2015 at 14:14
  • 2
    Some downtimes are not planned....
    – enderland
    Dec 5, 2015 at 14:14
  • 23
    Let me get this straight, it takes 16 minutes from downtime begin to an update that already identifies the issue, on a Saturday, and you're complaining it's too slow?
    – balpha StaffMod
    Dec 5, 2015 at 14:19
  • 3
    @balpha: Someone pressed the "make site read-only button". I'm asking that they (a) also take the ten seconds required to send a tweet "something's wrong ... stand by", or (b) remove the text from the read-only banner that tells me to check Twitter.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:20
  • 11
    Nobody pressed a button. Read-only mode enables itself automatically.
    – balpha StaffMod
    Dec 5, 2015 at 14:20
  • 1
    @enderland: When they occur, it is commonplace to make a page that says "service status" indicate that service is down. Or, don't have it.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:21
  • 3
    @balpha: Did not know that. Then it shouldn't instruct me to visit a status blog that hasn't been updated in three months. Or the blog should have an automatic health indicator powered by the same technology that auto-enables read-only mode. Y'know, that kind of thing.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:21
  • 4
    Well, it will be updated as soon as they know what's going on. I don't see what else you'd like to have. What's the point of tweeting that the site is down if we can all see that the site is down? How will it help to get everyone who hasn't noticed to come flocking to the site and trying to connect because they saw a tweet? Come on. You can wait ten minutes!
    – terdon
    Dec 5, 2015 at 14:23
  • 4
    @terdon: The point is knowing that it's an unexpected problem, vs unannounced maintenance. It's annoying when as far as you know the SE team just took the site down without warning. It's annoying when you have just written an answer and your finger's on the "Post" button, and you have no idea when you'll be able to submit it. Say there's a problem you're investigating and everything's fine. I'm not asking for miracles. In my line of work when one of my servers has a problem, I make sure the Service Status page has a big red icon to tell people. Whether I have diagnosed it yet or not.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:25
  • 1
    Sigh. Never mind.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:39
  • 4
    I agree with this to the extent of "the read-only-mode auto-danger-switch should put something on the info pages linked to in the banner; if not, then don't link to them". I can't really see manual Twittering being a priority when there's an emergency outage, though, however annoying it is for me as a user.
    – jscs
    Dec 5, 2015 at 19:21
  • 1
    @balpha what? So no big red button "READ ONLY MODE" which requires two passwords?
    – Shadow The Spring Wizard
    Dec 5, 2015 at 20:13
  • 4
    For the curious, here's what the top of the read-only control screen looks like.
    – Nick Craver
    Dec 5, 2015 at 20:26
  • 2
    @ShadowWizard ...has sensitive data :) Not passwords, but sharing lots of connection string data has no positive impact and doesn't add anything so...meh?
    – Nick Craver
    Dec 6, 2015 at 13:12

1 Answer 1

Reset to default
39

You seem to be asking Stack Exchange to prioritize telling everyone about the problem over fixing the problem.

As a sysadmin, this strikes me as entirely backward.

After all, you were already aware that there was a problem, based on the read-only message, the 500 errors, and the site being truly offline.

What we should be doing is resolving the problem as quickly as possible, and telling people about it when time permits. If this means that a tweet goes out five minutes later, so be it.

Improve this answer
9
  • 1
    Yes, in the situation where a problem was resolved in a couple minutes, I believe that is exactly what this question is asking for: a tweet in 5 minutes confirming that there was a problem on SE's end and saying that it was resolved. Or if even that's too much to ask, there's the last paragraph about removing the links to the blog and the Twitter account.
    – David Z
    Dec 5, 2015 at 14:17
  • 3
    It takes all of ten seconds to send a tweet. Anyone familiar with the art of prioritisation knows that this fact makes it worthwhile to send. It is more important than ever when there are problems/urgent activities going on that communication be maintained. The status page does not exist to inform us that everything is fine. Reminds me of something someone said to me at work the other day, with a project going mental leading to "we'll sort out the version control branches later". "We'll version these builds later." No: extra important during panics to stick to protocol!
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:17
  • 2
    "After all, you were already aware that there was a problem, based on the read-only message, the 500 errors, and the site being truly offline." No, see, this is the problem. I was not. Since there was no information provided and read-only mode usually comes up during maintenance, and since the Status blog is obviously abandoned, for all I knew it was unannounced maintenance (wouldn't be the first time). Lack of communication means your users/customers don't know what's going on.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:22
  • 2
    @LightnessRacesinOrbit First, the SRE team has to become aware of a problem and figure out what's going on. Only when they know there is a problem can they tell anyone about it.
    – Michael Hampton
    Dec 5, 2015 at 14:24
  • 6
    @LightnessRacesinOrbit what difference does it make? The point is that when the site is down, you can't access it. What do you care at that moment whether it is down for maintenance, because of a DDoS attack or because New York was hit by a meteorite? The site is down regardless. I'd like to see a post mortem but a while-mortem seems irrelevant.
    – terdon
    Dec 5, 2015 at 14:25
  • 5
    @MichaelHampton: No, not really. balpha says read-only mode is automatic when a fault develops. So, have a tweet get sent out by the same mechanism. Have a traffic light system on the service status blog, powered by the same mechanism. This is basic service support stuff.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:27
  • 5
    @terdon: You're suggesting I "just shouldn't care". Well then why bother with the service status page and status Twitter feed at all? Remove it (which I did suggest as a possible alternative). Either have it, or don't have it. But don't have it, fail to use it, then say "why did you care enough to want to read it?" Because that's nonsense.
    – Lightness Races in Orbit
    Dec 5, 2015 at 14:28
  • 1
    @LightnessRacesinOrbit That seems possible. Perhaps you should add that suggestion to your question, where it will get seen by people who are smart and don't read the comments?
    – Michael Hampton
    Dec 5, 2015 at 14:29
  • 2
    @LightnessRacesinOrbit It's a great way to let your users know what went down. That doesn't mean it needs to be prescient or work the instant something bad happens. The tweet went out pretty fast and the blog page will be updated with the post mortem. I just don't get why you would whine about a few minutes which I'm sure they spent running around trying to fix things so that you and I could enjoy the site. At ~8 AM on a Saturday, no less!
    – terdon
    Dec 5, 2015 at 14:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .