We are piloting a v11.2.1/Exchange 2010 iApp deployment (we currently run v10.2 in production) and have noticed a problem with OWA 2010. Most of the time, when behind the 11.2.1 load balancers, when a user clicks on the “Options” button in the upper-right hand part of OWA, the link does not work. The browser continues to spin, but the menu never appears.
We have not been able to reproduce this behavior on our 10.2.1. platform.
Hi Josh, which version of the Exchange iApp are you using? Have you taken a client side capture of the HTTP traffic? If so, can you see which request is getting hung up?
Can you post a sanitzed copy of your combined persistence iRule here?
We are running f5.microsoft_exchange_2010_cas.2012_06_08.
I have not taken a client side capture yet, but I can do that if we need to.
Here is a copy of the iRule:
when HTTP_REQUEST {
OAB and Autodiscover do not require persistence.
switch -glob -- [string tolower [HTTP::path]] {
"/microsoft-server-activesync" {
Direct all ActiveSync clients to a common pool; use Auth
header value if it exists (Basic auth only, which is the
default); otherwise we fall back to client IP
if { [HTTP::header exists "APM_session"] } {
persist uie [HTTP::header "APM_session"] 7200
} elseif { [HTTP::header exists "Authorization"] } {
persist uie [HTTP::header "Authorization"] 7200
} else {
persist source_addr
}
pool FC-EX2010-DC1-CAS_as_pool
COMPRESS::disable
CACHE::disable
return
}
"/owa*" {
Outlook Web Access
if { [HTTP::header exists "APM_session"] } {
persist uie [HTTP::header "APM_session"] 7200
} else {
persist cookie insert
}
pool FC-EX2010-DC1-CAS_owa_pool
return
}
"/ecp*" {
Exchange Control Panel.
if { [HTTP::header exists "APM_session"] } {
persist uie [HTTP::header "APM_session"] 7200
} else {
persist cookie insert
}
pool FC-EX2010-DC1-CAS_owa_pool
return
}
"/ews*" {
Exchange Web Services.
if { [HTTP::header exists "APM_session"] } {
persist uie [HTTP::header "APM_session"] 7200
} else {
persist source_addr
}
pool FC-EX2010-DC1-CAS_oa_pool
COMPRESS::disable
CACHE::disable
return
}
"/oab*" {
Offline Address Book. Persistence is not required for OAB
pool FC-EX2010-DC1-CAS_oa_pool
return
}
"/rpc/rpcproxy.dll" {
if { [HTTP::header exists "APM_session"] } {
persist uie [HTTP::header "APM_session"] 7200
} elseif { [string tolower [HTTP::header "Authorization"]] starts_with "basic" } {
persist uie [HTTP::header "Authorization"] 7200
} else {
persist source_addr
}
Finally, this assigns the Outlook Anywhere pool.
pool FC-EX2010-DC1-CAS_oa_pool
If the HTTP profile assigned to your virtual server enables Compression
or RAM Cache, you should un-comment either or both of the following lines
to turn off the respective feature(s) for Outlook Anywhere connections.
If you are using Web Accelerator, un-comment both lines.
COMPRESS::disable
CACHE::disable
return
}
"/autodiscover*" {
Requests for Autodiscovery information. No Persistence.
pool FC-EX2010-DC1-CAS_ad_pool
return
}
default {
This final section takes all traffic that has not otherwise
been accounted for and sends it to the pool for Outlook Web App
if { [HTTP::header exists "APM_session"] } {
persist uie [HTTP::header "APM_session"] 7200
} else {
persist source_addr
}
pool FC-EX2010-DC1-CAS_owa_pool
}
}
}
when HTTP_RESPONSE {
if { [string tolower [HTTP::header values "WWW-Authenticate"]] contains "negotiate"} {
ONECONNECT::reuse disable
ONECONNECT::detach disable
NTLM::disable
}
if {[HTTP::header exists "Transfer-Encoding"]} {
HTTP::payload rechunk
}
}
There are a few post-config steps that need to be done, like changing any “persist cookie insert” statements in the iRule to “persist cookie insert timeout 0”. This is due to a bug where the BIG-IP cookie has an expiration time, even though the cookie persistence profile indicates that it should be a session cookie. If your cookie has timed out, the next request might go to a different CAS and cause problems.
There’s a release candidate version of the iApp that takes care of all these post-config steps: DevCentral
Is it possible to redeploy using this version of the template?
We recommend doing a parallel deployment. The only thing that will conflict is the VIP address. You can create a new instance with the new template using a different IP address for the virtual server, then switch DNS to point to that IP; or you can reconfigure the existing deployment with a different IP, then create a new instance using the production IP address. Then you can delete the old instance.
Running into a problem. I run two instances of the iApp - one for CAS services, and one just for MAPI/RPC services. The CAS instance works fine until I configure a new instance just for MAPI. Then, the CAS HTTPS related services become unavailable.
It seems to maybe be related to both iApp instances using the same SNAT pool (More than 6000 clients)? But, this worked fine in earlier versions of the iApp.
Actually, the newer version of the iApp uses unique object names for the SNAT pools. Do you have the same addresses in those pools?
When you say that the HTTP services become unavailable, do you mean that the pool members are being marked down? Or are they healthy, but clients just can’t connect to them?
Which monitor type are you using, simple or advanced? Can you compare the monitor send/receive strings from both the new and old deployments? Are they continuously marked down, or are they flapping?
Something odd with the SNAT pools. Currently, I only have the CAS iAPP instance deployed. As soon as I assign the SNAT pool, the HTTP services become unavailable - they aren’t being marked down, but clients cannot connect to them. I do see this in /var/log/ltm:
Apr 25 10:18:26 slot1/tmm1 err tmm1[10353]: 01220001:3: TCL error: /Common/FC-EX2010-DC1-CAS.app/FC-EX2010-DC1-CAS_snatpool_irule3 - can’t read “static::snat_ips_FC”: no such variable while executing “snat $static::snat_ips_FC-EX2010-DC1-CAS([expr {[crc32 [IP::client_addr]] % [array size static::snat_ips_FC-EX2010-DC1-CAS]}])”
when RULE_INIT {
Use a local array to configure SNAT addresses.
These addresses do not need to be defined in a SNAT pool or elsewhere.
set static::snat_ips_FC-EX2010-DC1-CAS(0) 172.26.92.166
set static::snat_ips_FC-EX2010-DC1-CAS(1) 172.26.92.167
}
when CLIENT_ACCEPTED {
Calculate the crc32 checksum of the client IP
Use the modulo of the checksum and number of SNAT IPs to pick a SNAT IP
snat $static::snat_ips_FC-EX2010-DC1-CAS([expr {[crc32 [IP::client_addr]] % [array size static::snat_ips_FC-EX2010-DC1-CAS]}])
}
I just tested against my environment and indeed, the rule fails when the iApp name (and consequently, the iRule variable name) has a dash in it. I recommend using underscores in the iApp name: “FC_EX2010_DC1_CAS”.
Let me know if that doesn’t work for you. I’ll file a product bug against this if it’s not a known limitation of iRules. If it is, we can work around it in the next release of the template.
The issue is with variable names with hyphens in Tcl. This page does a pretty good job showing it:
The iApp could escape the hyphens using ${var-name} instead of $var-name. But it’s probably easier to just disallow the hyphen in the iApp name or replace them with underscores.