I’m on version 11.2.1 and using f5.microsoft_exchange_2010_cas.2012_06_08.
I created an “APM will provided secure remote access” iapp for one of our customers and it worked for the most part ( where you define the AD server, it does not respect routing domains, but not getting into that now as I can work around it)
I decided to create another one for a completely different customer and when I did it messed up the redirection on the first customer even though I shared no configuration. I can provide a lot more details but wanted to see if this is a known issue before I do as I really need to break the production customer and trace to page.
Steps taken:
1. Create first exchange iApp (working)
2. Create second Exchange iApp (First one now does not auto-login and I am dumped at the logon page for OWA)
3. Delete second Exchange iApp (First one now gets a “page cannot be display” with clearly an incorrect redirect)
4. Create a duplicate of the first Exchange iApp (even though this is a different VS, now the first iApp is working properly)
Now I did this just to confirm the issue.
5. Delete duplicate Exchange iApp (back to “Page cannot be displayed” on the original iApp)
6. Recreate the duplicate iApp. (Original working fine and this is where I sit today)
Activesync was fine throughout all of this, seems like purely a redirection/sso issue. I have no idea if the second iApp ever worked.
Any ideas,
Thanks in advance,
Misty
hi Misty, I think this may be due to a lack of uniqueness in the iRule that chooses an SSO configuration based on request URI. Can you post the text from the iRule(s) that end in “_select_sso_irule”?
thanks
Mike
Here is the one from the orginal:
when RULE_INIT {
set static::OWA_FORM_BASE_SSO_CFG_NAME “/vpn/wud_owa.app/exchange_forms_sso”
}
when ACCESS_ACL_ALLOWED {
set req_uri [HTTP::uri]
if { $req_uri contains “/owa/&reason=0” } {
WEBSSO::select $static::OWA_FORM_BASE_SSO_CFG_NAME
}
unset req_uri
}
This is the Duplicate one (remember I can’t delete this without the original one failing)
when RULE_INIT {
set static::OWA_FORM_BASE_SSO_CFG_NAME “/vpn/wud_owa_test.app/exchange_forms_sso”
}
when ACCESS_ACL_ALLOWED {
set req_uri [HTTP::uri]
if { $req_uri contains “/owa/&reason=0” } {
WEBSSO::select $static::OWA_FORM_BASE_SSO_CFG_NAME
}
unset req_uri
}
Thanks in advance,
Misty
Let’s test by changing “static::OWA_FORM_BASE_SSO_CFG_NAME” to “static::OWA_FORM_BASE_SSO_CFG_NAME_TEST” in *only* the wud_owa_test.app iRule:
when RULE_INIT {
set static::OWA_FORM_BASE_SSO_CFG_NAME_TEST “/vpn/wud_owa_test.app/exchange_forms_sso”
}
when ACCESS_ACL_ALLOWED {
set req_uri [HTTP::uri]
if { $req_uri contains “/owa/&reason=0” } {
WEBSSO::select $static::OWA_FORM_BASE_SSO_CFG_NAME_TEST
}
unset req_uri
}
You will need to disable strictness from the properties page of the wud_owa_test.app iApp first, then edit the iRule.
Every time you do a new test, you should delete any existing APM sessions for the test client first by clicking on Access Policy>Manage Sessions.
May I ask what the expected result of this test should be? To see if the original iApp keeps working? I just want to know what to look for.
Thanks,
Misty
Yes, the goal is for both iApps to work correctly. My thought is that changing the static variable name to something unique may prevent the iRule from selecting the wrong SSO configuration.
ok made that change
The original is working fine.
The duplicate SSO never worked, it dumps you at the outlook login page and making that change had no effect on that. Remember I only created the duplicate to fix the original. I have no idea why I thought that would work but it did.
Sorry, I intended that you would compare the iRules created for the 1st and 2nd customers, not the duplicate of the 1st iApp. At any rate, a good way to troubleshoot is to go to System>Logs>Configuration>Options and set the APM SSO log to debug level. Then, ssh into the BIG-IP and run a “tail -f /var/log/apm” command while the error is happening. F5 support can assist with this if necessary.
I appreciate the help. When I had the second customer’s config on there it broke the first production customer. I deleted it in desperation to try to fix the production customer (which it made the problem worst).
I don’t have the resources for setting up a lab. I was just hoping someone here had run into the issue before or could look at the iApp and see why they would step on each other. Support is going to assist me tomorrow on some Firepass to APM issues and this is on the list. Hopefully they can advise me.
I just wanted to bump this. After a long time of this being open with F5 support I was essentially told that the iApp doesn’t support multiple exchange servers configurations and I could try to build it myself or pro service consultants.
If it doesn’t work it shouldn’t let you even try to create a second iApp so you don’t break your production/working environment like I did. I’m really surprised this doesn’t come up more often. I can see all kinds of reasons to create a second even without multiple customers.
So if anyone ever run into this and figures out how to fix it please post. I pretty much have to leave those customers on Firepass until its fixed or we just take F5 out of the picture and put these servers in a DMZ.
Can you please send me your F5 case number?
I believe it should work with some post-configuration steps depending on the what you are doing with APM, but it will take me a bit of time to investigate that.
One issue I know of off the top of my head is the DNS server settings in the System menu. You can put multiple DNS servers in the list, but APM will only check the first one and if no information is found at that server for the requested domain, it considers that a successful response and doesn’t check any further down the list.
So, if your APM policies use objects that need to do DNS lookups and you have multiple policies doing lookups against multiple DNS servers that don’t know about each other, you are going to have problems with the policy that uses the 2nd DNS server on the list. An example of this is the AAA server object in v11.2 and earlier, which allows you to configure just an FQDN for the AAA server, which of course means that BIG-IP has to go look that IP up using DNS. In 11.3, the AAA server pool requires that you put in both a domain controller FQDN and IP address, so it doesn’t need to rely on DNS anymore. It’s a simlar story for the Kerberos SSO, which relies on reverse DNS lookups to build a Kerberos ticket request. This can also be worked around in BIG-IP.
I’ll do a bit of work here and let you know how it goes. It may require an upgrade to v11.3; looks like you are on v11.2?
Mike
I am on 11.2.1 HF5. An upgrade to 11.3 is planned soon.
Our DNS is good, though it is different agencies, DNS is central and under our control. I have confirmed this through packet captures and remember *only* the SSO on Outlook web access is broken. Active-sync continued to run flawlessly.
It looks like when you add the second iApp that it garbles the uri of the first iApp SSO. In fact, if you happen to know the correct uri and paste it in once you log in, webmail come right up. Its really specific. I think you were on to something with your first posts but I’m am not knowledgeable enough to fix it once the damage has been done. (other than deleting and recreating)
Misty,
This morning I successfully tested deploying two separate Exchange environments behind a single BIG-IP running APM. I can think of at least two features required by this solution that are only available in BIG-IP v11.3 and above, which are the AAA server pool and client-initiated forms SSO.
You will also want to use the latest version of the iApp, RC3, which we expect to release in the next day or so. Can you send me a private message on DevCentral with your email contact info? I can let you know as soon as that RC has been declared.
Mike
We’re running LTM 11.2.0 HF7 and see a similar issue, except that we do not use APM. The client connections terminate directly on the LTM’s virtual server.
We have 2 sets of Exchange 2010 CAS servers in two separate RDs. One pair of CAS servers were deployed last year with the iApp f5.microsoft_exchange_2010_cas.2012_06_08 and had been working fine.
Yesterday, we added a new instance to operate against a new pair of CAS servers in a different RD. The moment its added, the older iApp stops working (ActiveSync, Outlook Anywhere and OWA). Each iApp uses one FQDN for all Exchange services, but they are different for each iApp instance.
During the problem, it looks like the SNAT pool was not being currently chosen and/or the RDs were not being honored. The older iApp’s CAS servers (RD X) were seeing inbound connections from IPs belonging to the newer iApp’s SNAT pool (RD Y)
We’ve opened a case with F5 support.
Mike - is there an updated iApp template you mentioned?
Thanks.
speedbird,
There is a new version of the iApp here: DevCentral
However, I don’t know that it will help with your issue. I’ll do some searching as well, but Support should be able to tell you if this is a known issue. Please keep us posted.
Mike
Mike,
Appreciate the help, I’ll let you know if F5 support helps.
Mike,
As suspected, our issue was the SNAT pool configuration. The iApp creates an iRule with the $Name chosen by us, but within the iRule, the array name still needs to be unique across the different iApps.
For example:
**********
when RULE_INIT {
set static::iAppFoo_snat_ips(0) 1.1.1.1%816
set static::iAppFoo_snat_ips(1) 1.1.1.2%816
}
when CLIENT_ACCEPTED {
snat $static::iAppFoo_snat_ips([expr {[crc32 [IP::client_addr]] % [array size static::iAppFoo_snat_ips]}])
}
**********
Maybe this could be enhanced in the next iApp version? For now, we’ve worked around the problem by manually renaming the values within the iRule.
Thanks for your help.
Thanks for the response. This problem has been fixed in the upcoming release.