Lack of help in forums - I really need help - nobody home?

[Ruthiness]

I have several issues and the more I look the more entangled things get.
I have not gotten ANY response from anyone all day on my issues.



PLEASE HELP ME SOMEONE??????

I have posted several config files and output of commands.

ISSUES ARE:

Very slow to send mails from clients using Scalix as their SMTP server
Very slow to receive mails at times

In general, the customer views Scalix as a pain in the ass slow piece of crap.. so .. it would BEHOOVE Scalix to help me with the configuration.

i have followed ALL wiki directions to the T

I also have noticed problems with postgres in pgstartup.log - I don't know if we care about this or not or if it is related. We don't use Mobile client.

*****************ANYONE???? ************************************

I have search until my eyes are bleeding in this forum for answers and that is why I am imploring someone to help because, as I have seen other similar problems, they don't seem to be EXACTLY our problem.

I need to know how to find any errors or evidence of the bottleneck in logs but I have NO IDEA where to look in the logs to see what is happening.

i have done a telnet SMTP AUTH conversation with the server (which is FAIRLY COMPLICATED) and I notice pretty good pauses during AUTH and during the actual mail submission. Where do I look to see if there are errors anywhere that relates to these issues? I have turned on SMTPD logging - nothing. I see some errors in catalina.out regarding JDBC - related????

I am starting to NOT LIKE SCALIX when I find that it has so many moving parts that make it difficult to troubleshoot like this.

Here is your chance Scalix to redeem yourselves. Please don't let me down any more than you have already. i really want to like Scalix but...

[heupink]

Maybe check your dns settings..? Sometimes when mail servers are slow, it's because dns resolving is slow..?

Did you install the sysstat utilities? With them, you can check what keeps a server buzy. I run it on all my servers. (sar gives a nice overview, with this i discovered that disk io was a bottleneck on my old machine)

Does top show any extraordinary buzy processes? Memory/disk usage?

[jaime.pinto]

We might be able to help you better if you post some information about your setup. For example:
1) hardware configuration: RAM, HD/raid, #of NICs, etc?
2) OS and scalix versions
3) DMZ or behind the firewall with 192.168.x.x IP?
4) DNS pipeline, correct forward/reverse resolution, correct MX entries
5) everything self-contained? Fetchmail? External ldap/AD? etc
6) Is some firewall or mail filter/appliance getting in the way? All needed ports open?
7) Any difference in performance when comparing SWA to a client application
scalix connector or IMAP/SMTP on the client side.
9) versions of Outlook, XP/Vista, etc?

You don't need to provide specific info of your site, generic "scalix.domain.com/123.123.123.123" will do.
With that in mind some of us will ask you to run some commands, provide some further information and soon you and your clients will be much happier. Scalix is a very complex system for one person alone to grasp everything in the early stages, so don't feel bad about it. Just be patient and don't give up.

[Ruthiness]

Maybe check your dns settings..? Sometimes when mail servers are slow, it's because dns resolving is slow..?

Did you install the sysstat utilities? With them, you can check what keeps a server buzy. I run it on all my servers. (sar gives a nice overview, with this i discovered that disk io was a bottleneck on my old machine)

Does top show any extraordinary buzy processes? Memory/disk usage?

so far top is showing mostly idle cpu. I don't have good utils to monitor the disk i/o yet. I will look for sysstat utils.

I don't *think* it is a DNS issue but there ARE issues with another domain which is the domain they moved *from* - in that they set a duplicate A record to the new mail server's IP address but they used a different hostname - I don't know if this is causing issues but it COULD be reverse DNS probs from this.

I need proof of this however so I can basically put it back into their laps. Here is what they did with the OLD DNS:

set MX to a new A record using ip address of new mail server on new domain but the A record uses different hostname.. ie:

old domain DNS (using made up ip and domain info)

A 66.78.225.23 mailpa.olddomain.com

MX mailpa.olddomain.com

New domain DNS

A 66.78.225.23 mailpa. newdomiain.com
MX mailpa.newdomain.com

So if you reverse lookup 66.78.225.23, you may sometimes have it resolve to mailpa.olddomain.com

Scalix is installed at mailpa.newdomain.com

How can I prove that this is causing problems, if it is? What log file might show this might be causing delays?

[Ruthiness]

We might be able to help you better if you post some information about your setup. For example:
1) hardware configuration: RAM, HD/raid, #of NICs, etc?
HP DL380 4GB RAM RAID (I think 0+1) on two 146gb SAS 10K drives, 2 NICs - second used for backups on diff subnet

2) OS and scalix versions

2.6.9-55.0.9.ELsmp #1 SMP Tue Sep 25 02:16:15 EDT 2007 x86_64 x86_64 x86_64 GNU/Linux

rpm -aq | grep scalix
scalix-tomcat-connector-11.2.0.52-1.rhel4
scalix-sac-11.2.0.52-1
scalix-platform-11.2.0.52-1
scalix-res-11.2.0.52-1
scalix-mobile-11.2.0.52-1
scalix-server-11.2.0.11121-1.rhel4
scalix-swa-11.2.0.52-1
scalix-postgres-11.2.0.52-1
scalix-tomcat-5.5.23-342
scalix-sis-11.2.0.52-1


3) DMZ or behind the firewall with 162.168.x.x IP?
behind firewall with NATed IP

4) DNS pipeline, correct forward/reverse resolution, correct MX entries

See msg above regarding DNS - I do not control DNS I just make recommendations to them

Relevant info from nslookup using a fake IP for example

mailpa.olddomain.com 66.225.38.16
mailpa.newdomain.com 66.225.38.16 <--- Scalix is physically here

MX at olddomain.com --> mailpa.olddomain.com
MX at newdomain.com --> mailpa.newdomain.com

How do I prove they should not do this?

I also got a weird response at one point when I tested reverse DNS. I got a missing .com on the domain for where the Scalix server lives.

In other words if I do an nslookup of the IP Address I get this:

> 66.225.38.16
Server: 66.54.175.122
Address: 66.54.175.122#53

Non-authoritative answer:
16.38.225.66.in-addr.arpa name = mailpa.newdomain. <-----------------???
16.38.225.66.in-addr.arpa name = mailpa.olddomain.com.

Did the DNS provider mess up the A record maybe?
I did a dig of the domain and the hostname and did not see the problem.
Do I just point this out to the DNS provider and they will know where to fix?



5) everything self-contained? Fetchmail? External ldap/AD? etc
All on one box

6) Is some firewall or mail filter/appliance getting in the way? All needed ports open?
How do I find this out? I believe all necessary ports are open. We only get *delays* not errors getting/sending mail

7) Any difference in performance when comparing SWA to a client application
Seems slow for both they tell me

scalix connector or IMAP/SMTP on the client side.
Not using connector - only IMAPS/SMTP. I think the problem is with SMTP part

9) versions of Outlook, XP/Vista, etc?

all version of Outlook, Thunderbird and OS it seems Not just Vista, not just Outlook

I asked the client to switch some users to the ISP's SMTP server for outgoing to see the difference. The problem also happens for people no matter where they are - office or home - in neither case are they using a VPN to connect so it is all over the Internet and doesn't matter if over ISP at office or home.

You don't need to provide specific info of your site, generic "scalix.domain.com/123.123.123.123" will do.
With that in mind some of us will ask you to run some commands, provide some further information and soon you and your clients will be much happier. Scalix is a very complex system for one person alone to grasp everything in the early stages, so don't feel bad about it. Just be patient and don't give up.

[Ruthiness]

One interesting this also.. for some reason the domain has authoritative with three yipes.com servers when I do an nslookup - set q=any. If I do a whois for the domain, the nameservers are different. Shouldn't the authoritative nameservers for whois match what you get in nslookup?

I'm a little rusty on DNS best practices but this seems like a mismatch.

[Ruthiness]

It's only on the reverse DNS that I get the yipes.com authoritative.

Anyone know how or why these nameservers are authoritative for the reverse IP?
It sort of resolves correctly except for the missing .com on the actual domain.. duh!

[Ruthiness]

bingo..

when I do a tracert to the IP address from my pc - I get the mailformed DNS entry of:

mailpa.newdomain <----- missing .com

Ok how do we solve this? Who is at fault? How do I show them?

[jaime.pinto]

When you own and manage your DNS entries things are much easier.
It does feel like DNS discrepancy, and the delay could be due to the fall over from primary to secondary DNS resolution.

You have to be systematic about this and separate the *Internal* from the *External* access, so that you can rule out the server as the problem.

1) get on the server itself if you can, (66.225.38.16 ), start Firefox and open SWA. You'll need graphics enabled. If you have to do this remotely be sure to be on the same 66.225.38.0 network. Send emails to yourself and time it. It should not take more than 30 seconds for it to show on the inbox.

2) configure an IMAP client on the same 66.225.38.0 network, and use 66.225.38.16 for the incoming and outgoing(SMTP) server, and do the same thing: send emails to yourself.

While doing this take a peek on the /var/log/maillog, /var/log/syslog and /var/op/scalix/xx/s/sys/log_somewhere for any error/strange messages. if you see already any name resolution references that could be your confirmation.

11.2.0 of the scalix server is a very stable engine, and solid as far as DNS requirements.

If everything feels fast enough with 1) and 2) only using IP than you ruled out the server.

3) do the same with FQDN. Scalix is very fussy about what goes into /etc/hosts internal DNS and external DNS.

4) if the internal part is OK move on to the external DNS and external clients. Check that out against the NATed addressed. If this address resolves to something different from the external FQDN than that may be your problem. You may need a dedicated NATed IP just for the scalix server.

But for now I think you should focus on the DNS. It has to be perfect even from before you start the installation of scalix.

[Ruthiness]

DNS reverse IP thing has been fixed.

Seems outgoing is still rather slow.

Here is what I have observed on Thunderbird/Vista sending out a 5mb file zipped as an attachment after this DNS problem was corrected.

Sending the file itself took about 5 minutes, then it sat for about another five minutes while it seemed to send the file a second time - the message said "Sending authenticate login information" and there was a progress bar that looked like it was sending the attachment again..

Any ideas?

Next I will do a test using a different smtp server to send the exact same file and I will time it compared to using Scalix as smtp server. That should highlight where the problem is and where it isn't...

Thanks for your help

[mikevl]

Hi

It seems that these issues may be related to the SMTPD issue.

Is it possible to disable (coment out) the Clamv entries for testing?

Then only one of two outcomes. It goes faster or not.

Mike

[jaime.pinto]

I'd suggest you first use SWA to send your email for testing purposes.
Then use WindowsXP with OL2003 and 11.2 connector to do another test.
And lastly use Thunderbird with IMAP/SMTP client.
Do the 3 above from *internal* network first, then and external.

Personally I don't trust Vista or OL 2007 for anything reliable. And you didn't say whether or not your client was on the same network of the server. That's important so that we can start to rule out or in some possible sources of the problem.

When we engage in troubleshooting mode we need to be very systematic, and only add elements to the equation that we are absolutely sure about, while removing all those we are not.

[Ruthiness]

Here is the problem I finally have zeroed in on.

When sending a 5mb file as attachment through the Scalix server SMTP, it takes DOUBLE the time that other SMTP servers do because it goes through a SECOND upload it seems while it is "sending login information" or "sending authentication login information" in another version of Thunderbird.

In either case, sending mail through the Scalix server seems to upload the file with attachment TWICE.

I sent the SAME file from the same machine through a different SMTP server and it did not do this.

Where is this configuration on the Scalix server that is causing this behavior?

I feel that this might be why the users perceive it to be slow. I will try to test also with Outlook.

[Ruthiness]

This test I have just done proves these things:

The upload of the 5mb file is just as fast with the Scalix SMTP server as it is with another SMTP server - so the throughput is not the issue.

I observed the same behavior from both Windows 2003 Server and Vista, so it is not a client OS issue.

The issue it appears to be that the Scalix server is requiring a second upload of the mail with the attachment for some kind of "authenticate login" sequence.

I need to authenticate with the other SMTP server as well so it is NOT the Auth config that is causing this.

It finishes sending the 5mb file in about 2.5 minutes and THEN it opens a new dialog window that says "sending authenticate login info" which takes another 2.5 minutes to complete.

The received file with attachment shows a receive time that is within 1 minute of the time I put into the test mail, but when sending through the Scalix server, it ties up the client for a full 5 minutes! Not good!

I do NOT get this second round of "authenticate login info" with the other SMTP server.

Phew.. and now we finally have found the right symptom - but to which knob in Scalix does it belong?

[Ruthiness]

Also, I don't want to involve the Outlook connector at this point - I want to just look at the SMTP features we are trying to use for outgoing mail.

while waiting for a response, I will send the same attachment via Webmail to get the timing and behavior as well.