Just a slight clarification on what Rich just said
1) If you installed PostGIS with CREATE EXTENSION -- postgis functions, views, and included tables (except custom entries in spatial_ref_sys) aren't backed up anyway regardless what you do. Only thing that is included in the backup is the command CREATE EXTENSION postgis; and custom spatial_ref_sys rows if you added any. So it actually always ends up picking whatever the default postgis (that listed in share/extensions/postgis.control) you have in your install. If for whatever reason, you have multiple versions of PostGIS installed in your PostgreSQL instance, and you want your restore to not use the default PostGIS version (that listed in postgis.control), Then you would need to do CREATE EXTENSION postgis version "2.1.8"; --for example And then restore your data non-cleanly. 2) This issue is only an issue if your postgis extension is not in the same schema as your data – so best practices screws you in this regard. The reason why it's not an issue with same schema is that restore always adds the schema of the table being restored in search path. 3) This issue only affects data where either constraints or indexes defined on the table utilize functions that utilize other PostGIS functions or other non-pgcatalog functions. So Darrel to answer your question. You do not need it for all 16 constraint functions, just the 4 I mentioned that call out to other functions. The reason for that is that on a table, the index functions and constraint functions, are always fully-qualified, so it's only when that full-qualified function then makes an unqualified call to another function that the issue ensues. This issue isn't isolated to raster restores, but it affects raster much more that geometry/geography proper, since raster utilizes constraint functions that call other postgis functions and these get triggered as part of the restore to validate the constraints. If you were to for example create an index with any postgis function that happens to call another function and your data happens not to be in the same schema as where postgis is installed, then you will have a restore problem. In case of indexes it's especially annoying because you don't notice it until you notice things are much slower. E.g. if you have a spatial index that uses ST_Transform (that's the common case I run into), this issue crops up. So I also out of habit set the search_path on ST_Transform as well. Thanks, Regina From: postgis-users [mailto:[email protected]] On Behalf Of Richard Greenwood Sent: Wednesday, November 11, 2015 5:41 PM To: PostGIS Users Discussion <[email protected]> Subject: Re: [postgis-users] restore problem Darrel, A couple people have already pointed out that creating the postgis extension in the target database before you restore is advisable. Going along with that I also recommend excluding postgis from your backup. Postgis is put in the "public" schema so I avoid putting anything else in the "public" schema and then exclude it from my backups. The backups are a little smaller, there are fewer notices when you restore, and upgrading the postgis versions is a breeze. Best of luck, Rich On Wed, Nov 11, 2015 at 2:56 PM, Darrel Maddy <[email protected] <mailto:[email protected]> > wrote: Dear Regina, Many thanks for this. A least now it looks like this is nothing I have done (my first assumption with all problems is that I did something wrong – this assumption is usually correct). I had seen that ticket (this was the chatter I referred to) but assumed with the Milestone set at 2.0.8 I was hoping this had been fixed by 2.2 – clearly I do not understand this terminology! I will attempt to use this workaround tomorrow – hopefully it will resolve the restore issue and I will be back tomorrow full of praise :) Santa is bringing me a copy of your “PostrgreSQL: Up and Running book”, so hopefully once I am good to go with the basics I can really start to see what this combo can do. Best wishes Darrel From: postgis-users [mailto:[email protected] <mailto:[email protected]> ] On Behalf Of Paragon Corporation Sent: 11 November 2015 21:43 To: 'PostGIS Users Discussion' <[email protected] <mailto:[email protected]> > Subject: Re: [postgis-users] restore problem Darrel, Sadly you are not the only one. I have the issue documented here: https://trac.osgeo.org/postgis/ticket/2485 and I haven't come up with a generic enough fix I feel comfortable including in PostGIS code. The work-around is to do this 1) Create your database and do CREATE EXTENSION postgis; 2) Then add search paths to the functions used in check constraints raster_constraint_info_regular_blocking _raster_constraint_nodata_values _raster_constraint_out_db _raster_constraint_pixel_types e.g. ALTER FUNCTION _raster_constraint_pixel_types(raster) SET search_path=pg_catalog,public,postgis; 3) Then restore your data. I'll try to write a full script up in the ticket Hope that helps, Regina From: postgis-users [mailto:[email protected]] On Behalf Of Darrel Maddy Sent: Wednesday, November 11, 2015 2:03 PM To: PostGIS Users Discussion <[email protected] <mailto:[email protected]> > Subject: Re: [postgis-users] restore problem OK can I come at this from a different direction? Are there any backup options I should be selecting to eliminate this problem? I am really struggling to understand why I cannot backup and restore a raster table – surely others have had this issue? At present while my databases are small (i.e. <13GB) I can live with rebuilding if things fall over but these tables will grow to TB sizes soon (if I deploy this for real) and I cannot contemplate having no reliable backup. I have been selecting only the default options for backup – should I be doing something differently? Could this be a problem with the raster tables themselves? They seem to work with my SQL queries (including putting the tiles together and viewing in QGIS) but is there anything I should be doing to them before I select backup? If the backup does not report errors does this mean there are no errors in the backup? I realise I have much to learn using postgres but I see little point in putting in that much effort if backup/restore does not work for me. I really do appreciate the helpful suggestions that have been made so far but are there any further suggestions? Darrel From: postgis-users [mailto:[email protected]] On Behalf Of Darrel Maddy Sent: 10 November, 2015 10:25 PM To: PostGIS Users Discussion <[email protected] <mailto:[email protected]> > Subject: Re: [postgis-users] restore problem The default appears to be custom. I did not change anything when I did this. I just accepted the defaults. Darrel From: postgis-users [mailto:[email protected]] On Behalf Of Bborie Park Sent: 10 November 2015 22:20 To: PostGIS Users Discussion <[email protected] <mailto:[email protected]> > Subject: Re: [postgis-users] restore problem Odd. Looking at your error message again, it looks like the constraint is already in place by the time the data is getting loaded... What are you restoring from? Tar, custom or text? On Tue, Nov 10, 2015 at 2:16 PM, Darrel Maddy <[email protected] <mailto:[email protected]> > wrote: Dear Bborie, Apologies, I was obviously not clear – I was not asking about the problem itself, I can accept this is a difficult problem to resolve. I was simply asking for instruction on how to use your work around. This you kindly provided. Alas, however, even with the boxes checked, I get the same error and the raster table (dems) is empty. Darrel From: postgis-users [mailto:[email protected] <mailto:[email protected]> ] On Behalf Of Bborie Park Sent: 10 November 2015 21:57 To: PostGIS Users Discussion <[email protected] <mailto:[email protected]> > Subject: Re: [postgis-users] restore problem Given that the core problem has to do with the restore process' search_path, no amount of advice will help you get around the restrictions in place with the restore process without jumping into that process. In pgAdmin, you can check the boxes "Pre-data" and "Data" of the "Restore Options #1" tab. This won't restore the post-data portion of the backup as restoring post-data should be causing your error. You could also try unchecking the box "Exit On Error" of the "Restore Options #2" tab. -bborie On Tue, Nov 10, 2015 at 1:29 PM, Darrel Maddy <[email protected] <mailto:[email protected]> > wrote: Dear Bborie, Yikes – that makes little sense to me. I am new to both postgres and postgis and would have hoped that backup and restore would have been simple push-button exercise! As I am using PgAdmin I have no idea how to follow your instructions :( Anyone care to give me an idiots guide? Sorry Darrel From: postgis-users [mailto:[email protected] <mailto:[email protected]> ] On Behalf Of Bborie Park Sent: 10 November 2015 21:22 To: PostGIS Users Discussion <[email protected] <mailto:[email protected]> > Subject: Re: [postgis-users] restore problem This problem arises because PostgreSQL sets the search path during the restore process. Unfortunately, the search path is quite restrictive. You should be able to do your restore in three sections using pg_restore's --section flag. The error looks like it has to be the post-data section, which you should output to a text file, change the "SET search_path" statements and then run through psql. -bborie On Tue, Nov 10, 2015 at 1:11 PM, Darrel Maddy <[email protected] <mailto:[email protected]> > wrote: Dear all, Like many here I suspect I want to keep a copy of my main databases held on my workstation, on my laptop. I used PgAdmin to backup the databases concerned (the backup files look about the size I was expecting) but restore on my laptop did not successfully complete. All my shp file tables restored without issues but my raster tables (these are tiled rasters) would not. Looking at the error in the restore window of PgAdmin this appears to be a problem with a function/constraint. pg_restore: processing data for table "dems" pg_restore: [archiver (db)] Error while PROCESSING TOC: pg_restore: [archiver (db)] Error from TOC entry 3459; 0 94054 TABLE DATA dems postgres pg_restore: [archiver (db)] COPY failed for table "dems": ERROR: function st_bandmetadata(public.raster, integer[]) does not exist LINE 1: SELECT array_agg(pixeltype)::text[] FROM st_bandmetadata($1... ^ HINT: No function matches the given name and argument types. You might need to add explicit type casts. QUERY: SELECT array_agg(pixeltype)::text[] FROM st_bandmetadata($1, ARRAY[]::int[]); CONTEXT: SQL function "_raster_constraint_pixel_types" during inlining COPY dems, line 1: "1 0100000100000000000000344000000000000034C0E0CCCCCCB1D517418066666692F80C41000000000000000000000000..." I did a quick search and found some old chatter on this issue but from the messages I read I would have expected this issue to have been cleared up in postgis 2 . Obviously I cannot contemplate having mission critical data in a database which does not backup/restore correctly so I am assuming there is a fix which avoids this issue or there is another way to make a copy of the database for transfer elsewhere? Once again any help would be gratefully received. Best wishes Darrel ps. Apologies for the string of questions I have asked lately – I will go silent once more shortly as I must move on to other things. _______________________________________________ postgis-users mailing list [email protected] <mailto:[email protected]> http://lists.osgeo.org/mailman/listinfo/postgis-users _______________________________________________ postgis-users mailing list [email protected] <mailto:[email protected]> http://lists.osgeo.org/mailman/listinfo/postgis-users _______________________________________________ postgis-users mailing list [email protected] <mailto:[email protected]> http://lists.osgeo.org/mailman/listinfo/postgis-users _______________________________________________ postgis-users mailing list [email protected] <mailto:[email protected]> http://lists.osgeo.org/mailman/listinfo/postgis-users -- Richard W. Greenwood, PLS www.greenwoodmap.com <http://www.greenwoodmap.com>
_______________________________________________ postgis-users mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/postgis-users
