I know this is not a sanctioned setup, but we've grown pretty dependent on it. We have a single Umbraco database, and each developer has their own localhost copy of the website's code, including the Umbraco files. Media files are merged by hand between machines. This setup works great for all new Umbraco nodes that are created.
The problem begins any time a change is made to an existing node (my go-to test case is changing the icon on a doc type). The row value gets entirely reverted in the database after an unknown period of time. Not sure if it's minutes or hours, but changes never seem to make it to the next day.
My hypothesis was that the XML cache on each developer's machine was re-writing existing database records with older cached data. So naturally, we disabled the XML cache on all of the various Umbraco installations. I've confirmed that none of them have App_Data\umbraco.config on their machines any more. The problem persists.
So it seems there is still caching and re-writing going on, but where is the caching happening? I can't figure it out. I'm hoping to find another Umbraco or IIS setting that might help with this issue.
Are you using any plugins (e.g., uSync)? That's where I'd start.
Also, Umbraco now uses something called flexible load balancing. Basically, any website that connects to the database is considered to be part of the same load balanced cluster.
Cache instructions are written to a table (can't remember the name of the table) and the other machines in the cluster run those instructions to update their caches (XML, Examine, and so on). Might be something weird going on there.
What database are you using (e.g., MySQL, SQL Server, SQL Server CE)? I have seen data loss issues with SQL Server CE.
Could it simply be that another person has the doctype open on their machine while somebody else is making a change, then they save without reloading the page (thereby overwriting the prior change)?
That's interesting what you mentioned about the flexible load balancing. I wonder if that can be disabled?
We are using SQL Server 2k12.
Another person having the doc type open is almost certainly not the case. It's happening with data types, and even deprecated doc types that no one messes with (I delete the entire doc type, and it comes back from the dead). And the particular doc type I'm testing with hasn't changed in ages.
My other thought was to disable IIS caching, although that doesn't seem likely to store instructions that would ultimately update the database. (But to be honest, I don't know much about it.)
Single shared Umbraco DB + multiple concurrent Umbraco IIS installations = database record changes keep getting reverted (u 7.4.3)
I know this is not a sanctioned setup, but we've grown pretty dependent on it. We have a single Umbraco database, and each developer has their own localhost copy of the website's code, including the Umbraco files. Media files are merged by hand between machines. This setup works great for all new Umbraco nodes that are created.
The problem begins any time a change is made to an existing node (my go-to test case is changing the icon on a doc type). The row value gets entirely reverted in the database after an unknown period of time. Not sure if it's minutes or hours, but changes never seem to make it to the next day.
My hypothesis was that the XML cache on each developer's machine was re-writing existing database records with older cached data. So naturally, we disabled the XML cache on all of the various Umbraco installations. I've confirmed that none of them have App_Data\umbraco.config on their machines any more. The problem persists.
So it seems there is still caching and re-writing going on, but where is the caching happening? I can't figure it out. I'm hoping to find another Umbraco or IIS setting that might help with this issue.
Does anyone have advice for what to try next?
Are you using any plugins (e.g., uSync)? That's where I'd start.
Also, Umbraco now uses something called flexible load balancing. Basically, any website that connects to the database is considered to be part of the same load balanced cluster.
Cache instructions are written to a table (can't remember the name of the table) and the other machines in the cluster run those instructions to update their caches (XML, Examine, and so on). Might be something weird going on there.
What database are you using (e.g., MySQL, SQL Server, SQL Server CE)? I have seen data loss issues with SQL Server CE.
Could it simply be that another person has the doctype open on their machine while somebody else is making a change, then they save without reloading the page (thereby overwriting the prior change)?
Thanks for the response.
Nope, no plugins.
That's interesting what you mentioned about the flexible load balancing. I wonder if that can be disabled?
We are using SQL Server 2k12.
Another person having the doc type open is almost certainly not the case. It's happening with data types, and even deprecated doc types that no one messes with (I delete the entire doc type, and it comes back from the dead). And the particular doc type I'm testing with hasn't changed in ages.
My other thought was to disable IIS caching, although that doesn't seem likely to store instructions that would ultimately update the database. (But to be honest, I don't know much about it.)
is working on a reply...