I have a problem on the two heaviest-used Samba servers in my company. They both are exhibiting the same problem, just in different ways. Stargazer is my "Main" File and LDAP directory server - it functions as the "PDC" for my network. It's running Red Hat Enterprise Linux "ES" 2.1, a recompiled custom Red Hat "ES" version of Samba 2.2.7 (to add LDAP functionality), and OpenLDAP 2.0.27. Valykyrie is my "Print and Backup" server. It's running Red Hat 8.0, a recompiled Red Hat 8 version of Samba 2.2.7 (as above), OpenLDAP 2.0.27 and CUPS 1.1.17. With the exceptions of configuration for Primary LDAP server/LDAP client, the machines are configured roughly the same. Both were constructed using the IDEALX.org Samba+LDAP HOWTO. And both run quite well - for a while. Then, they will quit responding to connection requests an even local logon requests. This doesn't happen at the same time on both machines - it appears that the failures are not connected in any way other than both system logs report "Too many open files". On Stargazer it seems to be associated with OpenLDAP (slapd) and Valykyrie's problems seem to stem from nscd. This is shown in the following snippets from each "/var/log/messages": STARGAZER: --------------------------------------------------------------- Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open /etc/hosts.allow: Too many open files Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open /etc/hosts.deny: Too many open files Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open /etc/hosts.allow: Too many open files Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open /etc/hosts.deny: Too many open files --------------------------------------------------------------- VALYKYRIE: --------------------------------------------------------------- Sep 2 03:25:38 valykyrie nscd: 718: while accepting connection: Too many open files --------------------------------------------------------------- The reason I post this inquiry here is that both machines are running Samba and as such both depend on these daemons for proper functioning. I am hoping that someone on the list will be able to give a pointer as to the cause of the problem. If I need to ask this elsewhere, please advise me. I have 5 other machines running Samba just not with the load of the these two. The other machines seem to be unaffected by this problem - as yet. Any help will be appreciated... -- Kevin L. Collins, MCSE Systems Manager Nesbitt Engineering, Inc.
Chances are, you have too many open files. These servers are running on Linux. Linux limits the number of file handlers on a system wide basis to the value set in the sys.fs.file-max sysctl. This can be changed by editing the value in /proc/sys/fs/file-max, or using the sysctl utility. This is a per system (kernel) limit. There is also a per-process limit, which is a bit harder to change. Hope this helps you. This limit is here to prevent a DoS attack on any process from eatting up all the systems memory. File descriptors take memory in the kernel, so too many can consume it all. Choose a value that is reasonable for your workload. On Tue, 2003-09-02 at 12:41, Collins, Kevin wrote:> I have a problem on the two heaviest-used Samba servers in my company. They > both are exhibiting the same problem, just in different ways. > > Stargazer is my "Main" File and LDAP directory server - it functions as the > "PDC" for my network. It's running Red Hat Enterprise Linux "ES" 2.1, a > recompiled custom Red Hat "ES" version of Samba 2.2.7 (to add LDAP > functionality), and OpenLDAP 2.0.27. > > Valykyrie is my "Print and Backup" server. It's running Red Hat 8.0, a > recompiled Red Hat 8 version of Samba 2.2.7 (as above), OpenLDAP 2.0.27 and > CUPS 1.1.17. > > With the exceptions of configuration for Primary LDAP server/LDAP client, > the machines are configured roughly the same. Both were constructed using > the IDEALX.org Samba+LDAP HOWTO. And both run quite well - for a while. > > Then, they will quit responding to connection requests an even local logon > requests. This doesn't happen at the same time on both machines - it > appears that the failures are not connected in any way other than both > system logs report "Too many open files". On Stargazer it seems to be > associated with OpenLDAP (slapd) and Valykyrie's problems seem to stem from > nscd. This is shown in the following snippets from each > "/var/log/messages": > > STARGAZER: > --------------------------------------------------------------- > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.allow: Too many open files > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.deny: Too many open files > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.allow: Too many open files > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.deny: Too many open files > --------------------------------------------------------------- > > VALYKYRIE: > --------------------------------------------------------------- > Sep 2 03:25:38 valykyrie nscd: 718: while accepting connection: Too many > open files > --------------------------------------------------------------- > > The reason I post this inquiry here is that both machines are running Samba > and as such both depend on these daemons for proper functioning. I am > hoping that someone on the list will be able to give a pointer as to the > cause of the problem. If I need to ask this elsewhere, please advise me. > > I have 5 other machines running Samba just not with the load of the these > two. The other machines seem to be unaffected by this problem - as yet. > > Any help will be appreciated... > -- > Kevin L. Collins, MCSE > Systems Manager > Nesbitt Engineering, Inc.-- Jerry Haltom Feedback Plus, Inc.
I have this problem with Solaris 8.0 and OpenLDAP 2.0.27 standalone (without samba). I have other configuration with Solaris 9.0 and OpenLDAP 2.1.19 with no problems. I think you need to upgrade your OpenLDAP server version. On Tue, 2003-09-02 at 12:41, Collins, Kevin wrote:> I have a problem on the two heaviest-used Samba servers in my company. They > both are exhibiting the same problem, just in different ways. > > Stargazer is my "Main" File and LDAP directory server - it functions as the > "PDC" for my network. It's running Red Hat Enterprise Linux "ES" 2.1, a > recompiled custom Red Hat "ES" version of Samba 2.2.7 (to add LDAP > functionality), and OpenLDAP 2.0.27. > > Valykyrie is my "Print and Backup" server. It's running Red Hat 8.0, a > recompiled Red Hat 8 version of Samba 2.2.7 (as above), OpenLDAP 2.0.27 and > CUPS 1.1.17. > > With the exceptions of configuration for Primary LDAP server/LDAP client, > the machines are configured roughly the same. Both were constructed using > the IDEALX.org Samba+LDAP HOWTO. And both run quite well - for a while. > > Then, they will quit responding to connection requests an even local logon > requests. This doesn't happen at the same time on both machines - it > appears that the failures are not connected in any way other than both > system logs report "Too many open files". On Stargazer it seems to be > associated with OpenLDAP (slapd) and Valykyrie's problems seem to stem from > nscd. This is shown in the following snippets from each > "/var/log/messages": > > STARGAZER: > --------------------------------------------------------------- > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.allow: Too many open files > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.deny: Too many open files > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.allow: Too many open files > Aug 24 20:20:00 stargazer slapd[32271]: warning: cannot open > /etc/hosts.deny: Too many open files > --------------------------------------------------------------- > > VALYKYRIE: > --------------------------------------------------------------- > Sep 2 03:25:38 valykyrie nscd: 718: while accepting connection: Too many > open files > --------------------------------------------------------------- > > The reason I post this inquiry here is that both machines are running Samba > and as such both depend on these daemons for proper functioning. I am > hoping that someone on the list will be able to give a pointer as to the > cause of the problem. If I need to ask this elsewhere, please advise me. > > I have 5 other machines running Samba just not with the load of the these > two. The other machines seem to be unaffected by this problem - as yet. > > Any help will be appreciated... > -- > Kevin L. Collins, MCSE > Systems Manager > Nesbitt Engineering, Inc. > -- > To unsubscribe from this list go to the following URL and read the > instructions: http://lists.samba.org/mailman/listinfo/samba