Alfresco Server configuration for production

cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
Active Member

Alfresco Server configuration for production

Jump to solution

Hi All,

Currently, I am using alfresco community 5.0d version in production server.

We have allocated 16GB RAM to the hosted instance.

Currently, I am facing the issue of slowness. Also, out of box login API is taking 11 to 12 seconds to produce alf ticket.

Current subsystem properties.

### Solr indexing ###
index.subsystem.name=solr4
dir.keystore=${dir.root}/keystore
solr.host=localhost
solr.port.ssl=8443

The total size of the alf_data folder is 449GB. And here I am attaching the document which describes alf_data size. 

So can someone please suggest me any reason for this slowness?

1 Solution

Accepted Solutions
Highlighted
Alfresco Employee

Re: Alfresco Server configuration for production

Jump to solution
9 Replies
Highlighted
Alfresco Employee

Re: Alfresco Server configuration for production

Jump to solution
Highlighted
Customer

Re: Alfresco Server configuration for production

Jump to solution

Hi:

I see that you delete a lot in your instance (due the size of contentstore.deleted). This may require lots of resources for SOLR. Some basics recommendations included in Nathan's article :

- Increase JVM size and monitorize (at least via OOTB Support Tools)

- Reindex SOLR (if you delete a lot, your indices may be fat of deleted documents which occupy more space and perform worse when searching)

- Separate SOLR instance (use dedicated resources for your index and search server)

Regards.

--C.

Highlighted
Alfresco Employee

Re: Alfresco Server configuration for production

Jump to solution

Cesar Capillas

I guess that 90% of "contentstore.deleted" are DELETED nodes that are out of Database and SOLR.

If your assumption is true, and trashcan have been never been purged, then they have a huge node just under Archive root. And this can be also the cause of the low performance.

Software Engineer in Alfresco Search Team.
Highlighted
Customer

Re: Alfresco Server configuration for production

Jump to solution

Hi Angel Borroy‌:

Let me clarify, in my assumption a huge contentstore.deleted (with a similar size to the contentstore) means that:

  • A lot of content was deleted from trashcan, and some content nodes may still live in the database during some period. Many deletions involved a lot of JVM and CPU resources for Alfresco repository, and surely for SOLR (who is a RAM and CPU eater). And so, my first and third recommendations, for a future with many deletions. The use case of Alfresco should consider what process or why they are deleting so much..
  • I do not know really know what's inside the thrashcan. I imagine that, as he is deleting so much, it may have a lot of content in thrashcan that is being purged manually because of the version of Alfresco (without thrashcan cleaner). No more.
  • Finally, if you do many deletions, you will have many deleted documents between your SOLR indices, or better said, many documents flagged as deleted, between your indices, but not deleted from disk. I am not completely sure if this applies on modern versions of SOLR, but sometime ago I read that these deleted documents degrade the search performance. Here's some reference from Elastic.

Lucene's Handling of Deleted Documents | Elastic 

And so, my second recommendation for a full reindex for getting a healthy index in both cores.

Regards.

--C.

Active Member

Re: Alfresco Server configuration for production

Jump to solution

Hi Angel Borroy

Thank you so much for your valuable reply.

When I remove all data from the trashcan and contentstore.deleted.

After that, we perform a full reindex which is suggested.

What happens on next day system was slow down. Even login API takes 11 seconds to respond.

if u can just share the reason behind this behavior then it would be a great help.

Thanks in advance.

Highlighted
Alfresco Employee

Re: Alfresco Server configuration for production

Jump to solution

There is no need to perform a full reindex after removing content from trashcan...

Anyway, a full reindex is a hard task for your system. The text from every document in the server is extracted by Tika and even some transformations to PDF can be taken in the process. Probably you can inspect your server to identify the process that is consuming CPU, but I guess that is Java.

My advice: Don't perform a full re-index if there is no good reason (as data/indexes corruption). 

Software Engineer in Alfresco Search Team.
Highlighted
Active Member

Re: Alfresco Server configuration for production

Jump to solution

Hi Angel Borroy

Thank you so much for the reply.

When I delete all data from contentstore .deleted, then still any reference to that data which is stored in DB?

If yes then which is steps to be taken to solve that issue.

Thank you.

Highlighted
Alfresco Employee

Re: Alfresco Server configuration for production

Jump to solution

If you deleted all data from "contentstore.deleted", then you have lost references in database (for sure).

Probably you can understand better Alfresco deletion cycle at https://github.com/keensoft/alfresco-deleted-content-store-cleaner

Software Engineer in Alfresco Search Team.
Highlighted
Senior Member II

Re: Alfresco Server configuration for production

Jump to solution

All the documents that are moved to contentstore.deleted are not referenced in the DB anymore. You can savely delete them. We use a cron-Job to delete the files and folders from contentstore.deleted after X days - matching our backup-cycle