Forum Discussion
Ken_B_50116
Cirrostratus
CirrostratusGTM health monitor fails on pool, but not on member.
I am running GTM 11.4.1 HF1.
If I assign the default HTTP health monitor to a GTM pool, the pool goes red. If I remove the pool health check and add the HTTP health monitor to the member that's in the pool, the pool goes green.
The GTM has no problem making a telnet connection on port 80 to the member. It can ping and tracert to the member without problem.
This behavior happens for any pool and member combination. Is this expected behavior? Why can I not set a pool health check and have it work correctly? I spend most of my time in LTM and am accustomed to LTM behavior.
thanks!
Cory_50405Noctilucent
Does the server respond as expected if you issue the curl command?
From bash shell:
curl http://serveripaddress
Ken_B_50116Yes. From the GTM, when I open the member server's IP address with curl, I get expected HTML from the default document.
shaggyNimbostratus
The results should be the same between the two. Have you tried capturing the network-behavior with tcpdump, and is this GTM part of a network of iquery-connected GTMs/LTMs?
- Ken_B_50116
1) I have not make any packet captures yet, as this hasn't been a critical issue.
2) This GTM does communicate with LTMs using iquery, however the GTM does not use that data. All server/member/node objects in the GTM are added manually.
- Ken_B_50116
I suppose it's good to know that this behavior is not expected. This might motivate me to open a case with F5 support to inquire about the issue. At least I know it's not working how it is supposed to. Much of the config in the GTM is not intuitive to me, so sometimes I'm never really sure how things are supposed to work.
- Cory_50405
Short answer is no, this is not expected behavior. A support case is your best bet.
MatthiasRitter_Hi,
I'm experienced the same problem. Did you solve this?
GTM marks down the Pool because the member (VIP) seems to be down, but the VIP in the menu: GSLB -> Servers -> "PoolServer1" -> Virtual Servers .. is still up. Seems to be a problem with the "http"-health-check??!?
- Ken_B_50116
I have not resolved this. I have not opened a support case yet. I have seen one other GTM problem with a health check not working correctly for a pool, however I think it might be resolved by HF5 for 11.4.1. So I am planning to install that as soon as management approves it. I need this patch for other bug fixes as well.
If HF5 does not resolve the issue, then I will pursue it with F5 support.
- shaggy
@ken b and @matthiasritter - i doubt this solves the pool-level monitor vs. pool-member-level monitor discrepancy, but it's worth examining when troubleshooting GTM monitors
- do you have multiple GTM data center objects configured?
- in your GTM/LTM iquery mesh, regardless of having the correct rules allowing iquery communication, do you have firewalls/ACLs separating GTM/LTM devices that are in the same data center?
- what does /var/log/gtm say about the pool-level and member-level monitors when they fail/pass?
- Ken_B_50116
- Yes, we have 2 data center objects.
- Yes.
- Unfortunately I have not been able to reproduce the problem with the non-production objects I have to work with.
stan_peachey_86Cirrus
I experienced similar issue of GTM pools being marked down when upgrading from 10.2.4 to 11.4.1 HF4 (HF5 was not out). I punted to previous version to investigate. All server/member/node objects in the GTM were added manually. I suspected something changed with the custom health monitors, but it looks like you had the same problem with default monitors. I do not currently have a GTM/LTM iquery mesh. I plan to configure iquery with our LTMs when I get the GTMs up to 11.x. All pools are green in 10.2.4, so firewall rules should be in place.