The size of the indexes on the archive table are <51mb in both cases. There is no difference here. I totally agree.
I think the reason you don't see a difference in size is because of how null values are stored, I think in 1 byte but haven't found a reference. So yes even if you remove 20 unused types you only remove 20 bytes which as you point out is nothing. But the extra columns still affect read and write performance. Write isn't a big big deal as we don't do lots of writes anyway. But we might do lots of reads depending on what we are doing with our station data and we probably are all running this on inexpensive slow hardware. In my case a RPi but a new one which isn't all that slow other than if you're comparing it to something else that's new. But, for instance it still cut the time to make the daily summiers by more than half. Again not like we do that often so not a huge deal. This is where the real change probably came from. I also changed the data types of the observations from double (8 bytes) to float (4 bytes). Mysql made the sqllite data type doubles instead of floats. I don't have REAL_AS_FLOAT set and that's my fault. I am going to move to FLOAT(n) and set the precision on the columns next which won't change the row length, as the columns are all still 4 bytes, but to make things easier when I use other applications against this data set. In my case the length of the data went from ~1.1 gb to <650mb in this case. It also reduced the size of the binlogs, which get purged anyway. It also reduced the size of the *ib* files. It cut the time to and size of dumping the table almost by half, I haven't tried restoring yet but expect the same. Queries run faster. In my opinion there are other reasons to trim the schema to fit your needs other than the size of the data file. But yes it's more work and that depends on how you use your data if it's worth it or not. Obviously I think it's worth it and YMMV. -dk On Friday, October 9, 2020 at 9:01:49 AM UTC-4 [email protected] wrote: > Trimming the schema does not make as big a difference in database size as > you might think. > > For example, using my own database of 1.4M rows, trimming the schema from > 48 observation types to 27, reduces the size from 268MB to 201MB. > > The reason is that most of the space is taken up by the indexes, not the > column data. > > -tk > > On Thu, Oct 8, 2020 at 8:02 PM d k <[email protected]> wrote: > >> Yup.. I just found that and was about to report back I was trying it that >> was it. Just restarted the test system to see if it went away. I think I >> got rid of all of them now. >> >> Gary you are the best. Thanks so much. >> >> On Thursday, October 8, 2020 at 10:54:27 PM UTC-4 gjr80 wrote: >> >>> Hi, >>> >>> First up, thank you for not posting images of text, it’s makes >>> reading/searching logs a real pain. >>> >>> The error is due to a skin trying to generate a plot that involves >>> extraTemp1 and from the short log extract I would guess that this is >>> from the Seasons skin. If you look in the Seasons skin config file >>> (skins/Seasons/skin.conf) under [ImageGenerator] you will find the daytemp, >>> weektemp, monthtemp and yeartemp plots use extraTemp1 (and extraTemp2 and >>> extraTemp3). Easiest fix is to comment out those plots, eg: >>> >>> # [[[daytemp]]] >>> # yscale = None, None, 0.5 >>> # [[[[extraTemp1]]]] >>> # [[[[extraTemp2]]]] >>> # [[[[extraTemp3]]]] >>> >>> Save skin.conf and the error should go away on the next report cycle. >>> >>> Gary >>> On Friday, 9 October 2020 at 12:29:14 UTC+10 [email protected] wrote: >>> >>>> I tried to post this as an image but it doesn't show. So here is the >>>> text. >>>> >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: Caught >>>> unrecoverable exception in generator 'weewx.imagegenerator.ImageGenerator' >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** extraTemp1 >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** Traceback (most recent call last): >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** File "/usr/share/weewx/weewx/reportengine.py", line 197, >>>> in >>>> run >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** obj.start() >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** File "/usr/share/weewx/weewx/reportengine.py", line 280, >>>> in >>>> start >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** self.run() >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** File "/usr/share/weewx/weewx/imagegenerator.py", line 41, >>>> in run >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** self.genImages(self.gen_ts) >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** File "/usr/share/weewx/weewx/imagegenerator.py", line 176, >>>> in genImages >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** start_vec_t, stop_vec_t ,data_vec_t = >>>> weewx.xtypes.get_series(var_type, >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** File "/usr/share/weewx/weewx/xtypes.py", line 91, in >>>> get_series >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** raise weewx.UnknownType(obs_type) >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** weewx.UnknownType: extraTemp1 >>>> Oct 8 20:03:19 prometis weewx[271870] ERROR weewx.reportengine: >>>> **** Generator terminated >>>> Oct 8 20:03:19 prometis weewx[271870] DEBUG weewx.reportengine: Report >>>> 'SmartphoneReport' not enabled. Skipping. >>>> >>>> >>>> On Thursday, October 8, 2020 at 9:39:14 PM UTC-4 Duane Kerzic wrote: >>>> >>>>> Hi, >>>>> >>>>> Thanks for all the help you provided last time around. Thanks in >>>>> advance this time for your help. >>>>> >>>>> I wanted to clean up weewx.archive table and make it a bit smaller. So >>>>> I deleted the columns I don't think I'll ever use. But now I'm getting >>>>> this >>>>> in the system log. >>>>> >>>>> >>>>> I'm guessing that extraTemp1 is coded into one of those files but I >>>>> haven't looked to find out yet. >>>>> >>>>> I've shortened the average row length of the archive table to 126 from >>>>> 217 bytes. Huge difference when you have 10 years of data. >>>>> >>>>> -dk >>>>> >>>> -- >> You received this message because you are subscribed to the Google Groups >> "weewx-user" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/weewx-user/02d0a56e-c9fc-4e48-a74a-cdb6291474bbn%40googlegroups.com >> >> <https://groups.google.com/d/msgid/weewx-user/02d0a56e-c9fc-4e48-a74a-cdb6291474bbn%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > -- You received this message because you are subscribed to the Google Groups "weewx-user" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/weewx-user/853cbe79-ee93-49b3-90c3-6a9a02dab1b7n%40googlegroups.com.
