September 19, 2014

Is Replication Slave ever going to catch up? When?!

If you ever had a replication slave that is severely behind, you probably noticed that it’s not catching up with a busy master at a steady pace. Instead, the “Seconds behind master” is going up and down so you can’t really tell whether the replica is catching up or not by looking at just few samples, unless these are spread apart. And even then you can’t tell at a glance when it is going to catch up.

Normally, the “severely behind” thing should not happen, but it does often happen in our consulting practice:

  • sometimes replication would break and then it needs to catch up after it is fixed,
  • other times new replication slave is built from a backup which is normally hours behind,
  • or, it could be that replication slave became too slow to catch up due to missing index

Whatever the case is, single question I am being asked by the customer every time this happens is this: When is the replica going to catch up?”

I used to tell them “I don’t know, it depends..” and indeed it is not an easy question to answer. There are few reasons catching up is so unstable:

  1. If you have restarted the server, or started a new one, caches are cold and there’s a lot of IO happening,
  2. Not all queries are created equal – some would run for seconds, while others can be instant,
  3. Batch jobs: some sites would run nightly tasks like building statistics tables or table checksum – these are usually very intense and cause slave to backup slightly.

I didn’t like my own answer to The question, so I decided to do something about it. And because I love awk, I did that something in awk:

I don’t know if this is ever going to become a part of a Percona Toolkit, however since it’s pretty much a one-liner, I just keep it in my snippets pool for easy copy’n’paste.

Here’s a piece of an output from a server that was almost 27 days behind just yesterday:

Of course, it is still not perfectly accurate and it does not account for any potential changes in queries, workload, warm-up, nor the time it takes to run the mysql cli, but it does give you an idea and direction that replication slave is going. Note, negative values mean replication isn’t catching up, but values themselves are mostly meaningless.

Here’s what the weird acronyms stand for:

  • s_behind – current Seconds_Behind_Master value
  • d_behind – number of days behind based on current s_behind
  • c_sec_s – how many seconds per second were caught up during last interval
  • eta_d – this is ETA based on last interval
  • O_c_sec_s – overall catch-up speed in seconds per second
  • O_eta_d – ETA based on overall catch-up speed (in days)
  • O_eta_h – same like previous but in hours

Let me know if you ever find this useful.

About Aurimas Mikalauskas

Aurimas joined Percona in 2006, a few months after Peter and Vadim founded the company. His primary focus is on high performance, but he also specializes in full text search, high availability, content caching techniques and MySQL data recovery.

Comments

  1. Table checksums should not add delay, because pt-table-checksum won’t do anything until replication is caught up.

  2. Michael Wehrle says:

    Can you trust “Seconds_Behind_Master” to be accurate? I have been rebuilding servers lately for a client, and typically by the time we get the new slave built, its between 24-36 hours behind, and something like a dozen binlogs “behind”. What I have found is that the “Seconds_Behind_Master” doesn’t even show non-zero until the slave is on the same binlog as the master. That doesn’t even account for the fact that its often very inaccurate.

    Also… 27 days behind? Why not rebuild the slave with a hot backup from the master?

  3. Michael Wehrle says:

    Also, the code is quite nice actually. Succint and useful, given the intent.

  4. Baron,

    normally – indeed pt-table-checksum wouldn’t do that. Interestingly though, the slave in the example above was actually a [3 days] delayed slave (using pt-slave-delay) but pt-table-checksum was configured to not check its status, for a pretty obvious reason that it will always be at least 3 days behind. We still want checksums to run though, to check that our last-resort backup is consistent.

  5. Michael, –

    it’s not completely accurate, but you also have to be sure you know what the number means. It is NOT “how long until the slave will catch up”, but rather “what is the difference between NOW() and the time when master was writing the events [I'm currently executing] into the binary log”. In that sense, it is accurate to one second.

    One problem with Seconds_Behind_Master is that if will show zero if it is IO thread, not SQL thread that is behind.

  6. Bill Willits says:

    Very nice post. And, super-useful. I’ve had the same need for this tool many times. Thanks for creating.

  7. aftab says:

    Would mk-slave-prefetch help in this situation?

  8. Thanks for sharing the script.

    The line with:
    grep Seconds_Behind_Master | awk ‘{print \$2}’

    Could be rewritten as:
    awk ‘/Seconds_Behind_Master/ { print \$2 }’

    It might be possible to use getline and make it more pure awk.
    http://www.gnu.org/software/gawk/manual/html_node/Getline_002fVariable_002fPipe.html#Getline_002fVariable_002fPipe

    AWK is a great tool. The only downside is that it’s probably not available by default on most Windows and other non-UNIX (like) operating systems.

    Rewriting this script to run server side (stored procedure/view/etc) seems favorable, but is not possible:
    http://code.openark.org/blog/mysql/reading-results-of-show-statements-on-server-side

  9. aftab, it might, though I’d probably try Yoshinori’s Replication Booster or Domas’ Replication prefetcher first, if need be.

  10. Daniël, –

    thanks for your tips!

  11. Thanks for sharing this script! It is quite useful!

  12. Frank Van Damme says:

    yes, pretty useful. A lot better than to keep running “show slave status” and making a guess about the time to catch up.

  13. Why not use statsD or some other data collector in a cronjob and just plot it on a graph? Lots of online tools that will capture/display this stat over time nicely for ya.

  14. Thanks for your comment, Ross. In this particular case I’m not interested in having the data plotted, I assume that most of the time (like, virtually ALL the time), slave is not behind the master and I ask this question (and use this script) very rarely when I get a task like setting up a new replication slave (or fixing a broken one) from a backup which is, say one week behind – and then we’re curious to know when we will have the slave up to date. This fix sometimes is an additional index, RAM increase or anything of that sorts.

    Note that as MySQL consultants (I’m not talking Remote DBA now) we work with variety of customers that would often come for a quick 16-32 hour gig and we don’t have any control nor time to have any graphing tools installed.

  15. Bill Willits says:

    Hey Aurimas.
    I have found this code to be very useful. The prediction of catch-up time is remarkably accurate!
    I created a script that add some bells and whistles – like notification via email during the catchup process and when fully caught up. The script relies on the heartbeat table for computing the lag so you need to have that.
    I hope it is useful to others. (and , I hope this pastebin works).
    http://pastebin.com/H2pxihrx

    ~Bill

  16. Bill, thanks so much! I do see the the pastebin just fine. I’m happy to hear that you find the catch-up time to be accurate, I also find that after 20-30 minutes or more (depending on a number of factors) it stabilises and shows pretty close ETA.

  17. Aurimas – makes sense. I stumbled on this post when trying to diagnose a slave not catching up. For others and I believe it may be covered in other posts here – setting innodb_flush_log_at_trx_commit=0 helped a permanently lagging slave catch up quickly. I then set it back to 1 once it caught up and its fine and not lagging. HTH this helps someone else.

    R

Speak Your Mind

*