gmane.linux.keepalived.devel

new keepalived release

Wolfram Schlich — 2008-11-24T15:15:01

Hi, when will the next keepalived version be released? I've read there are quite some patches to be incorporated... Thanks.

keepalived VRRP part used for N+1 redundantservices

2008-11-13T13:00:02

Hello, I would like to use just the VRRP part of keepalived to setup an N+1 redundancy scheme for a service that runs on the VRRP hosts. Just to give a simple example: - 2 active web servers, one on IP 1.2.3.4 and on IP 1.2.3.5 - 1 backup web server, that can take over one (or both) of the IP addresses given above (and will also take over the WEB service) This is how the configuration file would look like on the backup server: ---- vrrp_instance VI_10 { state BACKUP interface eth0 virtual_router_id 10 priority 50 authentication { auth_type PASS auth_pass password } virtual_ipaddress { 1.2.3.4/24 brd 10.0.11.255 dev eth0 } vrrp_instance VI_11 { state BACKUP interface eth0 virtual_router_id 11 priority 40 authentication { auth_type PASS auth_pass password } virtual_ipaddress { 1.2.3.5/23 brd 10.0.11.255 dev eth0 } ---- So BOTH VRRP INSTANCES WOULD BE RUNNING ON THE SAME ETHERNET INTERFACE (eth0). If I try to configure it like this, keepalived complains about a bogus VRRP packet,

keepalived_healthcheckers flooding syslog

Jack Neely — 2008-11-11T22:33:19

Folks, I'm migrating keepalived 1.1.15 from old servers running RHEL 4 to new servers running RHEL 5. I've had no problems moving VIPs between the different platforms, in fact its worked quite smoothly. However, the healthchecker thread seems to flood syslog on the RHEL 5 servers. Every delay_loop interval I get a log from each of the three RS that happen to be down: Nov 11 17:29:36 lvs00 Keepalived_healthcheckers: Timeout connect, timeout server [xxx.xxx.xxx.aaa:7777]. Nov 11 17:29:36 lvs00 Keepalived_healthcheckers: Timeout connect, timeout server [xxx.xxx.xxx.bbb:80]. Nov 11 17:29:36 lvs00 Keepalived_healthcheckers: Timeout connect, timeout server [xxx.xxx.xxx.ccc:80]. On the RHEL 4 platform I only got one syslog message when a host failed its healthcheck and it was silent until the RS came back online. Any idea what the difference may be? Jack Neely

keepalived - virtual_routes allow/enable tablexyz option

Pieter Smit — 2008-11-07T19:46:30

------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Keepalived-devel mailing list Keepalived-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/keepalived-devel

vrrp_sync_group weigth

MarisRuskulis — 2008-10-30T09:45:03

Hello! I havea bit unclear about vrrp_sync_group and track weights. I have two vrrp instances - httpd and mysql both in sync group, if one service fail both vips move to backup. I dont understand how to add track scripts to those instances, because track weights are ignored because of sync group. ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Keepalived-devel mailing list Keepalived-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/keepalived-devel

is keepalived still under development ?

Jérôme Loyet — 2008-10-27T23:07:14

bug: keepalived fails over when time is adjusted

Dominik — 2008-10-22T08:58:32

Hi I sent this email from another email account already, but somehow it wasnt received it seems. Therefor I send it again. I apologize if it should be sent and received twice. I use keepalived 1.1.15 on debian linux 2.6.18-5-686. When time is adjusted on the backup machine (e.g. via netdate/ntpdate or "date -s"), and this adjustment exceeds advert_int times 3 seconds in the future, the backup takes over the master role as it apparently did not receive the master's vrrp packets for the last $seconds. Easily testable with date -s "10 seconds" on the slave. In the log, you will see something like: Oct 22 00:03:55 dktest2sles10 Keepalived_vrrp: VRRP_Instance(VI_1) Transition to MASTER STATE Oct 22 00:03:55 dktest2sles10 Keepalived_vrrp: VRRP_Instance(VI_1) Received higher prio advert Oct 22 00:03:55 dktest2sles10 Keepalived_vrrp: VRRP_Instance(VI_1) Entering BACKUP STATE Depending on the speed and load of the machines, IP addresses even make it to the backup machine and so the backup machine

Strange problem

Vik Nat — 2008-09-30T16:37:33

Helo All, I changed my config to use DR , where I have setup a VIP 192.168.100.xx and realservers on the same subnet. Realservers are using IIS, so have setup the loopback adapters on both machines and am able to connect to the VIP on port 80 when I connect a client to 192.168.100.xx network. I add a DNAT using iptables to NAT the external IP address to the VIP and add a rule to allow access to port 80: iptables -t nat -A PREROUTING -p tcp -m tcp -d 66.xx.xx.xx --dport 80 -j DNAT --to-destination 192.168.100.xx (VIP in keepalived) iptables -A OUTPUT -p tcp -m tcp -d 192.168.100.xx --dport 80 -m state --state NEW iptables -A FORWARD -p tcp -m tcp -d 192.168.100.xx --dport 80 -m state --state NEW Now I go ahead and connect the client machine to the external interface and try and estaiblish a telnet session using port 80 using the external IP that I setup in iptables, but it does not connect and I see this in the logs: Sep 30 11:42:31 FW01 kernel: [67146.597410] ip_rt_bug: 66.xx.xx.xx -> 1

Fw: Help with Load Balancing

Vik Nat — 2008-09-25T19:37:51

Oh one thing I forgot to add that as soon as I add the NAT/Policy Rules, I am also unable to connect to VIP. ----- Forwarded Message ---- From: Vik Nat yahoo.com> To: keepalived-devel< at >lists.sourceforge.net Sent: Thursday, September 25, 2008 3:34:59 PM Subject: Re: [Keepalived-devel] Help with Load Balancing Ok - So right now I have the LB on port 80 as well as VRRP(Active/Passive) working. For the load balancing , I used an external IP as the VIP ( 66.xxx.xxx.xxx) and NATed to the internal addresses of the real servers. I can now connect to the VIP using a client on the outside and have tested the VIP by unplugging each real server from the swatch to see if LB is working as expected. So far so good on this . I have also tested the director VRRP failover and that works fine too. Now I need to configure iptables to NAT all the other machines that are not going to be loadbalanced and add policies for restricting services. As soon as I add the NAT and the rules, the VRRP gets hosed and bot

keepalived not working

Tears ! — 2008-09-18T10:23:33

sorry servers, keepalived bugs?

Jack Neely — 2008-09-16T18:26:47

Folks, In my experimentation with setting up a sorry server I've run across a few...more interesting things. My virtual servers normally look like this: virtual_server fwmark 27 { delay_loop 7 lvs_sched wrr lvs_method DR persistence_timeout 3600 protocol TCP real_server xxx.xxx.xxx.234 { weight 100 HTTP_GET { url { status_code 200 path / } connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } } Normally having more than one real_server block of course. Adding sorry_server xxx.xxx.xxx.154 directly below the "protocol TCP" line causes keepalived's health checkers to go crazy. It logs as fast as possible bringing up the load on the machine to about 2. Sep 16 13:38:09 hostname Keepalived_healthcheckers: Netlink reflector reports IP xxx.xxx.xxx.10 added Sep 16 13:38:09 hostname Keepalived_healthcheckers: Netlink reflector reports IP xxx.xxx.xxx.100 add

Help with Load Balancing

Vik Nat — 2008-09-15T14:30:48

Hello All, I have setup LVS/Keepalived on Ubuntu Server 8.04. VRRP works as expected but the LB on port 80 is not working. I can ping the real servers from the Master Director and telnet to them to port 80. I can ping the VIP ( 192.XXX.XXX.22) as well from the Director but cannot telnet into it on port 80, I get connection refused. The Realservers are both Windows, running IIS. I have setup the loopback adapters on both the real servers using the VIP. There is currently no FW loaded on both Directors. Any help or guidance would be much appreciated Thanks, Vik Here is my configuration: MASTER: global_defs { lvs_id lvs01 } vrrp_sync_group G1 { # must be before vrrp_instance declaration group { VI_1 VI_2 VI_3 } } vrrp_instance VI_1 { state MASTER interface eth0 virtual_router_id 51 priority 100 authentication { auth_type PASS auth_pass somepw } virtual_ipaddress { 66.XX.XXX.200/27 dev eth0 } } vrrp_instance VI_2 { state MAS

DR, Sorry Server, Best Practices

Jack Neely — 2008-09-11T18:03:55

Folks, I run a fairly large collection of VIPs through keepalived using LVS-DR. I need to setup a generic sorry server (simple web page) that I can include in the virtual_server configuration block for our web services. What I don't want to do is have to add VIPs or otherwise alter the configuration of the sorry server itself every time we use it with a different web pool. I've got 160 some odd VIPs so you can see why I want to avoid this step. :-) Do folks have sorry servers setup in similar situations? How did you handle the configuration? Does the transparent proxy method (the REDIRECT target in iptables) still work with RHEL 5? Thanks! Jack Neely

DR and iptables FORWARD question.

MarisRuskulis — 2008-09-11T09:15:06

Hello! So I have a trivial question. I set up Keepalived VRRP and LVS via Direct Routing and iptables for firewalling - this all on balancer. Then I needed to configure iptables FORWARD table, but what I found is - there is no traffic on FORWARD, this is what confuses me! How LVS forwards traffic? ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Keepalived-devel mailing list Keepalived-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/keepalived-devel

Can I disable fixed IP addresses on the network where keepalived shares VIPs?

Brian Ghidinelli — 2008-09-09T01:48:54

I think the answer to this question is yes but I wanted to ask in case I'm missing some unforeseen consequence. My public network is currently: xx.xx.xx.123 = lvs1 xx.xx.xx.124 = VIP for public website xx.xx.xx.125 = lvs2 My vrrp config looks like: vrrp_instance INET { state MASTER interface eth0 virtual_router_id 51 priority 100 debug 0 advert_int 5 lvs_sync_daemon_interface eth4 virtual_ipaddress { xx.xx.xx.124/28 brd xx.xx.xx.127 } } I want to remove lvs1/lvs2 from the public network so the only IP address accessible to the Internet is the shared VIP .124. If I disable the .123 and .125 addresses, can lvs1/lvs2 still successfully share the .124 address over eth0? Thanks, Brian ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to a

ipvsadm issue

alessio — 2008-08-26T08:16:36

Hi. I want to configure a virtual server with 2 real servers. But... I want to contact a real server via NAT and the other via Direct Routing. In keepalived the lb_kind option is "global" for the entire virtual_server; there is no way to specify the kind of LB for each real server? Like: virtual_server 10.10.10.2 389 { delay_loop 6 lb_algo rr protocol TCP real_server 192.168.192.3 389 { lb_kind NAT TCP_CHECK { connect_timeout 3 } } real_server 10.10.10.10 389 { lb_kind DR TCP_CHECK { connect_timeout 3 } } } I've noted that using ipvsadm it is possible to do such thing. And it works. ipvsadm -A -t 10.10.10.2:389 -s rr ipvsadm -a -t 10.10.10.2:389 -r 10.10.10.10:389 -g ipvsadm -a -t 10.10.10.2:389 -r 192.168.192.3:389 -m I've tried to use notify_down and notify_up in the keepalived real_server configuration, but without success... Any hint? (I'm using keepalived 1.1.13) Ciao, a --------------------------------------

Getting state informations on keepalived

2008-08-17T22:05:32

Dear list members, I successfully configured keepalive to get vrrp transitions to work. Now I'm looking for a possibility to check, which of the nodes is currently master and which slave. Is there a snmp interface, a state table that could be used, ... to get this information? I'm only (at the moment) interested in the vrrp part (not lvs) because I need this for firewalls to takeover all VIPs of all interfaces. Any hints how to get this information are highly appreciated. Thank you very much. Best regards Christof ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Problem: virtual_server_group to reduce health check traffic in a big installation - alternatives?

Thorsten Grote — 2008-07-31T07:09:30

Hi, I'm using a keepalived with fwmarks. As I used my real server not only for one pool but for many, I've quite a lot health checks to the same real server from different pools - that is where I liked to use virtual_server_group's. Heres is the problem: My fwmarks pools are overlapping for some reason, like this: virtual_server fwmark 1: R1 R2 R3 virtual_server fwmark 2: R2 R3 R4 virtual_server fwmark 3: R1 R4 R5 Ok, now you might wonder about the reason. The reasons is, that the application served in pool 1 might go crazy (to much traffic, mistakes, what ever reason). Therefor, I don't want to take a risk and put all my users/cutomers in only one pool. Further, I don't want to have the real servers exclusive for each customer, to not waste computing power when a customer is just idle. As a result, I'm overlapping the pools, which is just fine and does what I want. Back to the problem: If I now start using virtual_server_group, I would have to do something like this: group 1: R1-R6 virtual_se

'track_script' doesnt seem to work?

Jeffrey 'jf' Lim — 2008-07-29T06:01:30

keepalived reload behaviour, try 2

Siim Põder (kodust — 2008-07-25T06:33:10

Hi I posted a patch yesterday and it seems I understood some of the code wrong. This is a new and simpler (and somewhat tested) patch to make sure IP addresses (and possibly routes) are brought up and down as expected on vrrp reload. It does the following: 1. Keeps states during reload: not entirely sure if it's needed, but seems to be the right thing to do anyways 2. Keeps vipset, so that keepalived would remove it's IPs on the next possible transition to BACKUP 3. After diff_vrrp_vroutes/vip that remove addresses that are no longer in the configuration, also re-adds all vips and vroutes, in case there are any new ones. AFAIK static addresses/routes should already be processed as expected (didn't test though). Is this a sensible way to do it? Siim ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for tw

patch fixing keepalived vrrp behaviour on reload

Siim Põder — 2008-07-23T15:20:12

Hi! I'm using keepalived to switch the vrrp master back and forth during hardware maintainances and such. For this purpose, I have script that essentially does: 1. Takes template configuration file 2. Replaces some tags (prio, routerid, state) 3. Propagates the new conf to both servers 4. Reloads (sighup) keepalived on both servers However, it turned out that state switches occuring because of the reload did not add/remove IP addresses as they were supposed to. I would have to manually clear IP addresses off the router that switched to BACKUP state. In order to fix this bug, I made 2 changes: 1. Upon reload, keep the old vrrp states. AFAIK there should be no need to forget the state on reload (MASTER stays MASTER and BACKUP stays BACKUP - any occuring election can change it if priorities have changed) 2. Upon reload, remember which IP addresses and routes where added to kernel (vrrp->vipset and ipaddr->set + iproute->set). This is the important bit, as otherways keepalived would just forget it ever adde