Dears
can Erpnext work in big companies?
many users and branches
and a lot of daily transaction
thanks a lot
In most cases your bottleneck is going to be the database. @hfarid has implemented ERPNext in a company with 30,000 employees, so its possible, but there will be a few things you will have to optimize
4 Likes
Sure it can work with big companies.
You have to do the following 
-
You have to increase the timeout for NGINX and supervisor this could be done by
bench config http_timeout
bench setup supervisor
bench setup nginx -
Some Huge transactions could be made using Triggers in Database or by using multi record update using MariaDB update statement.
-
You have to update the configuration for MariaDB (my.cnf) and change the appropriate parameters for big queries.
And you will run smoothly for big company.
Best luck.
9 Likes
@h_farid_gm Would you be willing to speak on this subject at a meetup or a community webinar? Your experience and use case seems like it would make for really good educational content.
3 Likes
600 sec for timeout should work?
You mean for innodb buffer size? Can you share configuration?
What about using stored procedures and/or FK and cascade updates and delete?
Do you think Frappe ORM could be improved in some to get better performance? As far as i know when document is deleted it enqueue a background job to check and delete every linked documents, could this be demanded to database using FK?
2 Likes
The my.cnf changed parameters are as follows but it depends on how much RAM you have in your server
this case is for a server with 16 GB RAM :
max_connections = 100
connect_timeout = 50
wait_timeout = 6000
max_allowed_packet = 256M
thread_cache_size = 128
sort_buffer_size = 64M
bulk_insert_buffer_size = 128M
tmp_table_size = 256M
max_heap_table_size = 256M
net_buffer_length = 100K
key_buffer_size = 512M
table_open_cache = 400
myisam_sort_buffer_size = 512M
concurrent_insert = 2
read_buffer_size = 64M
read_rnd_buffer_size = 32M
query_cache_limit = 8K
query_cache_size = 256M
slow_query_log_file = /var/log/mysql/mariadb-slow.log
long_query_time = 10
* InnoDB
innodb_log_file_size = 1G
innodb_buffer_pool_size = 2G
innodb_log_buffer_size = 8M
innodb_file_per_table = 1
innodb_open_files = 400
innodb_io_capacity = 400
innodb_flush_method = O_DIRECT
12 Likes
For the multi records update I prefer doing it through python using frappe.db.sql function.
You can use stored procedure or triggers also.
3 Likes
@hfarid: How do you handle with the multi-branch permission? For example, a branch A only see the record and statistics from branch A, not branch B. Or the case of the warehouse manager in 1 warehouse can make some record on other warehouse even they don’t have the permission.
1 Like
You can do that from user permissions.
2 Likes
@hfarid Would you be willing to do a wiki post/guide about this? I imagine the post could cover the scale (of the implementation) you required and the changes (to settings) you have made, the infrastructure you are using and a rough estimate of costs and maintenance the system requires. This would be extremely useful for a lot of our users.
2 Likes
We just offload our entire DB to Amazon RDS that allows us to scale very easily.
4 Likes
Sorry, should have been more specific. It’s the cloud database service provided by Amazon :
Also, our servers are right on EC2 and they’re on the same region, so there’s almost no latency.
@vjFaLk Oh i know about that service , it seems kind of hard to estimate monthly cost for it …isn’t it?
Here’s the pricing :
