antirez 4267 days ago. 252242 views.
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).

I think this is big news for users as instead to see us developers and other Redis experts handwaving about the amount of copy-on-write being proportional to number of write ops per second and time used to produce the RDB or AOF file, now they get a number :-)

# How it is obtained?

We use the /proc/<pid>/smaps, so yes, this is Linux only.
Basically it is the sum of all the Private_Dirty entries in this file for the child process (actually you could measure it on the parent side and it is the same).

I verified that the number we obtain actually corresponds very well with the physical amount of memory consumed during a save, in different conditions, so I'm very confident we provide an accurate information.

# Why a number in the log file instead of an entry in the INFO output?

Because even before calling wait3() from the parent, as long as the child exits we no longer have this information. So to display this information in INFO we need some inter process communication to move this info from the child to the parent. Not rocket science but for now I avoided adding extra complexity. The current patch is trivial enough that we could backport it into 2.6 for the joy of many users:


The log is produced at NOTICE level (so it is displayed by default).
🚀 Dear reader, the first six chapters of my AI sci-fi novel, WOHPE, are now available as a free eBook. Click here to get it.
blog comments powered by Disqus