Skip to content

389 ds ignores nsslapd-maxdescriptors #6010

Closed
@resposit

Description

@resposit

Issue Description
389 ds 2.3.6 seems to ignore nsslapd-maxdescriptors=32768

Package Version and Platform:

  • Platform: RockyLinux 9.3
  • Package and version: 389-ds-base-2.3.6-3.el9_3.x86_64

Steps to Reproduce
In /etc/systemd/system/dirsrv@cloud20.service.d/override.conf I have:

[Service]
LimitNOFILE=32768
LimitNPROC=16384

In dse.ldif I have:

nsslapd-maxdescriptors: 32768
nsslapd-reservedescriptors: 16384

When start directory server:
systemctl start dirsrv@cloud20.service

I get the error:
[12/Dec/2023:13:00:09.358667991 +0100] - ERR - slapd_daemon - Not enough descriptors to accept any connections. This may be because the maxdescriptors configuration directive is too small, the hard limit on descriptors is too small (see limit(1)), or the reservedescriptors configuration directive is too large. Try increasing the number of descriptors available to the slapd process. The current value is 16384. 16384 descriptors are currently reserved for internal slapd use, so the total number of descriptors available to the process must be greater than 16384.
It seems 389 ds ignores the setting nsslapd-maxdescriptors=32768 and considers nsslapd-maxdescriptors=16384. In fact I change in dse.ldif:

nsslapd-maxdescriptors: 32768
nsslapd-reservedescriptors: 8192

Then the server starts properly. Strangely it then reports the correct nsslapd-maxdescriptors value:

[root@cloud20 etc]# dsconf -D "cn=Directory Manager" ldap://localhost config get nsslapd-maxdescriptors
Enter password for cn=Directory Manager on ldap://localhost:
nsslapd-maxdescriptors: 32768

Additional context
I have a twin server (same configuration) running 389-ds-base-2.2.9-2.el8.x86_64 and I have no such issue

Metadata

Metadata

Assignees

Labels

needs triageThe issue will be triaged during scrum

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions