HUGE database size

[willisadams]

We started with Alfresco Community Edition 5.0.d.  We later upgraded to the 201508 Early Access release of 5.1.  Later still, we upgraded to the 201509 and now, today, the 201510 EA.

The updates all went smoothly. I followed the directions found here: https://alpha-labs.net/2015/02/upgrading-alfresco/

During the latest update (from 201509 to 201510), the pg_dump step took a lot longer than it usually does.  When it finished, the dump file size was almost 16GB! Our entire alf_data directory only has around 65,000 files totaling 12.6GB.

Still, I loaded dump file into the new installation and everything seems to be working okay.  Using the pgAdmin tool I can connect the running alfresco database and see that it has a total size of 9275MB.  All of the tables in the database are pretty small (a few dozen KB or MB) with the exception of two tables: alf_prop_serializable_value is 8297MB and alf_prop_value is 872MB.

Is this normal?  If not, any suggestions on getting this problem under control?

Thanks in advance for your assistance.

EDIT: For reference, we are uploading most of our files to Alfresco using the built-in SMB server, but we're also using the Share web interface and the built-in IMAP server (clients use Thunderbird to connect). We aren't using any third-party add-ons. Our OS is CentOS 7 running in a VM on a Debian 8 KVM host. Let me know if you need any other info or want me to post logs.

[afaust]

That sounds as if the property cleaner actually did delete a substantial amount of entries from the database, despite all the errors in the log. The VACUUM only reclaimed the space freed by the cleaner. Can you please check the size of the two tables in the coming days and report back if they change substantially?

Regards
Axel

[willisadams]

Sorry for not following up on this sooner!  I just got an email from JIRA about this on some updated comments, and it looks like it's been marked resolved as of December.

Anyway, I'm happy to report that that database size has remained relatively small over the past few months.  We've got 115,000 items in our contentstore folder totalling 22 GB and the database size is just over 1 GB.  The alf_prop_value and alf_prop_serializable_value tables are very small, 152 KB and 32 KB, respectively. 

The largest tables are alf_node (387 MB) and alf_node_properties (342 MB), alf_child_assoc (115 MB), alf_transaction (60 MB), alf_node_aspects (55 MB), alf_content_url (45 MB), and alf_content_data (38 MB).  All the other tables are a few hundred KB or less.  I assume this is normal. 

Since the original post, we've upgraded to the 201512 EA release, the 201602, 201604, and 201605 General Releases, and we are now using the the 201606 EA Early Access release.  Everything seems to working well.  Aside from the occasional transformation failure, the log files are clean and our users haven't reported any problems. 

Thanks again for your help, and thanks to the Alfresco developers for putting together a great product!