[Ceilometer] To survey how to improve the performance of OpenStack Ceilometer

Frankly speaking, OpenStack Ceilometer will suffer some kind of performance issues sooner or later if you don't modify or tune the configuration. The issues has two parts that need you to consider. One is the message bus and API loading, and the other is database. However, I find some best practices which are easy and quick for us to adopt. Here you go:

1. Telemetry(Ceilometer) best practices

http://docs.openstack.org/admin-guide/telemetry-best-practices.html

a. Data collection

1. Based on your needs, you can edit the pipeline.yaml configuration file to include a selected number of meters while disregarding the rest.
2. By default, Telemetry service polls the service APIs every 10 minutes. You can change the polling interval on a per meter basis by editing the pipeline.yaml configuration file.

   for example:
   
   
   vim /etc/ceilometer/ceilometer.conf
   
   => evaluation_interval=120
   
   
   
   
   vim /etc/ceilometer/pipeline.yaml
   
   => interval: 120
   
   

   
   
3. 
   
4. you can delay or adjust polling requests by enabling the jitter support. This adds a random delay on how the polling agents send requests to the service APIs. To enable jitter, set shuffle_time_before_polling_task in the ceilometer.conf configuration file to an integer greater than 0.

b. Data storage

1. We recommend that you avoid open-ended queries.
2. You can install the API behind mod_wsgi, as it provides more settings to tweak, likethreads and processes in case of WSGIDaemon. a. For more information on how to configure mod_wsgi, see the Telemetry Install Documentation.
3. The collection service provided by the Telemetry project is not intended to be an archival service. Set a Time to Live (TTL) value to expire data and minimize the database size.
   

   for example:
   
   
   
   vi /etc/ceilometer/ceilometer.conf
   
   => time_to_live=302400
   
   
   

4. Use replica sets in MongoDB. Replica sets provide high availability through automatic failover. If your primary node fails, MongoDB will elect a secondary node to replace the primary node, and your cluster will remain functional.
1. For more information on replica sets, see the MongoDB replica sets docs.
5. Use sharding in MongoDB. Sharding helps in storing data records across multiple machines and is the MongoDB’s approach to meet the demands of data growth.

2. Metering Service (Ceilometer): Best Practices and Optimization

http://docs.hpcloud.com/#commercial/GA1/ceilometer/1.1commerical.services-reporting-bestpractice.html

a. Modifying the List of Meters

sources:
    - name: meter_source
      interval: 604800
      meters:
          - "instance"
          - "image"
          - "image.size"
          - "image.upload"
          - "image.delete"
          - "volume"
          - "volume.size"
          - "snapshot"
          - "snapshot.size"
          - "ip.floating"
          - "network.*"
          - "compute.instance.create.end"
          - "compute.instance.delete.end"
          - "compute.instance.update"
          - "compute.instance.exists"
      sinks:
          - meter_sink
sinks:
    - name: meter_sink
      transformers:
      publishers:
          - notifier://

b. Modifying the Polling Intervals

The interval attribute is the time between polls. Meters that are available as both notification and polling are going to be polled at the specified interval. To rely on notifications rather than polling, set the interval attribute to 604800 seconds, or once a week.

Reference

http://docs.openstack.org/developer/ceilometer/architecture.html

https://www.linkedin.com/pulse/dive-openstack-ceilometer-mishael-wexler

http://redhatstackblog.redhat.com/2016/01/18/ceilometer-polling-performance-improvement/

One of the main issues operators relayed was the polling that Ceilometer was running on Nova to gather instance information. It had a highly negative impact on the Nova API CPU usage, as it retrieves all the information about instances on regular intervals.

[Linux] Why does Linux require moving IP from eth interface to bridge interface?

This could be a common problem if you have KVM ( or other hypervisor in Linux ) on your physical server and want to use bridge mode with your VMs. At the same time, you also want to let your physical server has the network that can be accessed from other hosts at the same network subnet. At this moment, when a network interface (e.g., eth0) is added to a Linux bridge (e.g., br0), the IP address must be removed from eth0 and added to br0 for the networking to function properly.

I find some answers as follows:
http://unix.stackexchange.com/questions/86056/why-does-linux-require-moving-ip-from-eth-interface-to-bridge-interface
http://unix.stackexchange.com/questions/52674/why-doesnt-eth0-get-an-ip-address-when-set-up-in-a-bridge

==>
The NIC represents the uplink cable. A cable is layer 1, not layer 3. Now the Bridge works as the device that is being addressed for network traffic (incoming) on the server - either on layer 2 (Ethernet/MAC) and/or layer 3 (IP). So the device that responds to ARP-requests is the bridge - which is good, since it needs to distribute the traffic to the other interfaces on that bridge. If the responding device were the NIC, traffic would not be passed further on to the bridge.

==>
Normally it does not make sense to put any L3 protocol address on port interfaces - because incoming packets are diverted to the bridge interface before the L3 protocol is examined. This means the L3 protocol running on the port interface will never see any incoming packets.

[Neutron] Neutron Cheat Cheat Sheet

In recent days, the hand drawing style is becoming more and more pervasive in Taiwan. I don't know how it happens, but at least I can draw the Neutron cheat cheat sheet to echo this kind of style for fun.

P.S: This picture is originally for my colleagues to trouble shooting the Neutron networking problems.

[Neutron and SDN] Warm-up for understaning the integration of Neutron and SDN

I just spend some time to study the integration of Neutron and SDN. Furthermore, I also take a look at how ODL and ONOS are integrated with OpenStack Neutron. The following content and picture ( with URL link) are excerpted from the variety of resource in the internet. Some of sections has my comment with P.S. I think it can give you a clear concept of Neutron and SDN controller.

Neutron and SDN

P.S: This picture gives an overall architecture about Neutron and SDN controller that are integrated together.

http://thenewstack.io/sdn-controllers-and-openstack-part1/

When an OpenStack user performs any networking related operation (create/update/delete/read on network, subnet and port resources) the typical flow would be as follows:

1. The user operation on the OpenStack dashboard (Horizon) will be translated into a corresponding networking API and sent to the Neutron server.
2. The Neutron server receives the request and passes the same to the configured plugin (assume ML2 is configured with an ODL mechanism driver and a VXLAN type driver).
3. The Neutron server/plugin will make the appropriate change to the DB.
4. The plugin will invoke the corresponding REST API to the SDN controller (assume an ODL).
5. ODL, upon receiving this request, may perform necessary changes to the network elements using any of the southbound plugins/protocols, such as OpenFlow, OVSDB or OF-Config.

We should note there exists different integration options with the SDN controller and OpenStack; for example:

1. one can completely eliminate RPC communications between the Neutron server and agents on the compute node, with the SDN controller being the sole entity managing the network
2. or the SDN controller manages only the physical switches, and the virtual switches can be managed from the Neutron server directly.

http://www.sdnlab.com/16364.html

Neutron底層網絡的兩種模型示意如下。

第一種模型a) Neutron相當於SDN控制器，plugin與agent間的通信機制（如rpc）就相當於簡單的南向協議。第二種模型中Neutron作為SDN應用，將業務需求告知SDN控制器，SDN控制器再通過五花八門的南向協議遠程控製網絡設備。

第二種模型b) 也可以把Neutron看做超級控制器或者網絡編排器，去完成OpenStack中網絡業務的集中分發。

Plugin Agent

P.S: This picture shows the process flow between agents, api server and ovs when creating a VM.
http://www.innervoice.in/blogs/wp-content/uploads/2015/03/Plugin-Agents.jpg

About ML2

https://wiki.openstack.org/wiki/Neutron/ML2/LenovoML2Mechanism

http://zhouyaguo.github.io/openstack/2015/11/26/neutron-plugin-arch/

Neutron plugin体系

How OpenDaylight to integrate with Neutron

http://cntn2014.co/n/7545/content/opendaylight-one-best-controllers-openstack-7545.html

http://events.linuxfoundation.org/sites/events/files/slides/ODL%20and%20OpenStack%20-%20Workshop_0.pdf

How is ONOS to integrate with Neutron

SONA Architecture

onos-networking plugin just forwards (or calls) REST calls from nova to ONOS, and OpenstackSwitching app receives the API calls and returns OK. Main functions to implement the virtual networks are handled in OpenstackSwitching application.

OpenstackSwitching (App on ONOS)

Neutron + SDN Controller (ONOS)  

P.S: ONOS provides its ONOSMechanismDriver instead of OpenvswitchMechanismDriver

Reference:
Here is an article to talk about writing a dummy mechanism driver to record variables and data in logs
http://blog.csdn.net/yanheven1/article/details/47357537

[Proxy ARP] What is Proxy ARP?

Why I mention Proxy ARP is because in OpenStack environment, this function exists in FIP namespace with DVR enabled. The way to check whether the FIP namespace's Proxy ARP enabled is here:
# ip netns exec fip-545b57e2-0fa5-46da-89a2-591f7a5474ce cat /proc/sys/net/ipv4/conf/fg-2f3f4992-23/proxy_arp
1

We also can see the arp and ip mapping:
# ip netns exec fip-545b57e2-0fa5-46da-89a2-591f7a5474ce ip neighbor
10.12.20.32 dev fpr-6ddabb95-1 lladdr 42:72:88:ca:07:72 STALE

The following is the explaination about Proxy ARP
http://linux-ip.net/html/ether-arp-proxy.html

Occasionally, an IP network must be split into separate segments. Proxy ARP can be used for increased control over packets exchanged between two hosts or to limit exposure between two hosts in a single IP network. The technique of proxy ARP is commonly used to interpose a device with higher layer functionality between two other hosts. From a practical standpoint, there is little difference between the functions of a packet-filtering bridge and a firewall performing proxy ARP. The manner by which the interposed device receives the packets, however, is tremendously different.

Example 2.10. Proxy ARP Network Diagram



The device performing proxy ARP (masq-gw) responds for all ARP queries on behalf of IPs reachable on interfaces other than the interface on which the query arrives.

FIXME; manual proxy ARP (see also Section 9.3, “Breaking a network in two with proxy ARP”), kernel proxy ARP, and the newly supported sysctl net/ipv4/conf/$DEV/medium_id.

For a brief description of the use of medium_id, see Julian's remarks.

FIXME; Kernel proxy ARP with the sysctl net/ipv4/conf/$DEV/proxy_arp.

[Docker] the first experience with building docker image

It is essential to look for the docker image first if you need some services or functions running on docker container. But, once you want to customize it, you probably need to build your own docker image. The official document gives you a very complete description for you to do so. Please refer to this https://docs.docker.com/engine/userguide/containers/dockerimages/
The following command list are my steps to build a customized Drupal docker image.
The are two ways to build your own image:

1. Updating and committing an image

First, it would be better to have a Docker Hub account like this:


Second, to create a repository for your docker image.


If it's done, you can see this:



So, we can continue to the next step.
# Download the offical Drupal docker image
$ docker search drupal
$ docker pull drupal
$ docker images

# Create a container and update it ( be aware of the follwing parameters )
$ docker run -i -t --name danny_drupal -p 8000:80 drupal /bin/bash
  -i, --interactive               Keep STDIN open even if not attached
  -t, --tty                       Allocate a pseudo-TTY
  -p, --publish=[]                Publish a container's port(s) to the host

# From now on, you can go anything for your container
root@2a0849519c71:/var/www/html# apt-get update
root@2a0849519c71:/var/www/html# apt-get install openssh-server cloud-init -y
root@2a0849519c71:/var/www/html# exit

# To commit my changes
$  docker commit -m "Added my services" -a "teyenliu" \
danny_drupal teyenliu/drupal:v1

# Have to login docker before you push your change
$ docker login --username=teyenliu
$ docker push teyenliu/drupal

# Now it is successful to push your own drupal image and you also can see it on docker hub:


# To test your own drupal image:
$ docker run --name danny_drupal -p 8000:80 -d teyenliu/drupal:v1

# To check if the container is running
$ docker ps

# Open your browser with http://127.0.0.1:8000

2. Buidling an image from Dockerfile

$ vim Dockerfile

# This is a comment

FROM drupal:latest

MAINTAINER TeYen Liu <teyen.liu@gmail.com>

RUN apt-get update && apt-get install -y git

RUN apt-get install -y openssh-server

RUN apt-get install -y cloud-init

$ docker build -t teyenliu/drupal:v2 .
$ docker push teyenliu/drupal:v2

# The drupal repository will append the image of tag:v2




P.S: If you want to put the docker image to OpenStack Glance for further using, here is an example command:
$ docker save teyenliu/drupal | glance image-create --container-format docker --disk-format raw --name teyenliu/druapl

[Python] Problem with Python logging RotatingFileHandler in Django website

If seeing the log files are not rotated properly or correctly in Django web site, you most likely encounter the problem as the following article described:

Problem with Python logging RotatingFileHandler in Django website

"The log is done via RotatingFileHandler which is configured with 10 log files, 1000000 byte each. The log system works, but this are the log files I get:

-rw-r--r-- 1 apache      apache          83 Jul 23 13:30 hr.log
-rw-r--r-- 1 apache      apache      446276 Jul 23 13:03 hr.log.1
-rw-r--r-- 1 apache      apache      999910 Jul 23 06:00 hr.log.10
-rw-r--r-- 1 apache      apache         415 Jul 23 16:24 hr.log.2
-rw-r--r-- 1 apache      apache      479636 Jul 23 16:03 hr.log.3
-rw-r--r-- 1 apache      apache         710 Jul 23 15:30 hr.log.4
-rw-r--r-- 1 apache      apache      892179 Jul 23 15:03 hr.log.5
-rw-r--r-- 1 apache      apache         166 Jul 23 14:30 hr.log.6
-rw-r--r-- 1 apache      apache      890769 Jul 23 14:03 hr.log.7
-rw-r--r-- 1 apache      apache      999977 Jul 23 12:30 hr.log.8
-rw-r--r-- 1 apache      apache      999961 Jul 23 08:01 hr.log.9

As you can see it is a mess. Last log has been written to file hr.log.2 (Jul 23 16:24) instead of hr.log"

I did some survey and found the root cause is in here:
RotatingFileHandler bugs/errors and a general logging question
"The logging system is thread-safe but not safe
against multiple processes (separate Python instances) writing to the
same file. It certainly sounds like you need a scalable solution - and
having each script send the events to a network logging server seems a
good way of handling the scalability requirement. "

    These words above remind me a lot about the importance of synchronization when using multi-threading and multi-process. Also, scaling is another important item that a lot of people don't care. I want to highlight that we should be vigilant about these.

So, it is not your fault. Don't blame yourself. ( Sorry, I am kidding you ... )
Here is one solution for this issue. Please check out the following link ( it's written in Chinese ):
http://www.djangochina.cn/forum.php?mod=viewthread&tid=118752

[Linux Bonding] 802.3ad bond interface has shown RX dropped packets

If someone uses Linux Bonding and finds some or a lot of RX dropped packets in bond interface, please ignore these dropped packet message because of the following informations:

1. Linux Bonding and Single Nic with 1GE switch are no difference in packet loss
I use iperf tool with UDP to test packet drop and jitter and the result shows that there are no difference of packet loss between linux bonding and single nic.




2. Bond0 RX packet dropped is not a bug.
Please check out this: https://bugs.launchpad.net/ubuntu/+source/bridge-utils/+bug/1041070

This is related to the bonding mode and _not_ a bug. The bonding module will drop duplicate frames received on inactive ports, which is normal behavior. [0] Overall the packets should be getting into the machine without problems since they are received on the active slave. To confirm this do the following

1) Check dropped packets from all interfaces. So if eth0/eth1 are connected to bond0, we may see dropped packets for bond0 and eth0, but not for eth1. This depends on which interface is the active interface. This can be checked using the following:
cat /sys/class/net/bond0/bonding/active_slave

So if the active_slave isn't dropping packets, and the inactive slave is dropping packets this is normal in 'active-backup' mode (or any mode where there is an inactive slave).

2) If we want both interfaces to not drop packets we can use 'all_slaves_active' bonding module parameter [0].
Check:
cat /sys/class/net/bond0/bonding/all_slaves_active, it should default to 0 which means drop frames on the inactive slave.

If we set this to 1, we will no longer drop frames:
echo 1 | sudo tee /sys/class/net/bond0/bonding/all_slaves_active

3. This article suggests to turn off rp_filter ( could reduce RX dropped )
https://platform9.com/support/neutron-prerequisites-linuxkvm-overlays-vxlangre-vlans/

echo net.ipv4.conf.all.rp_filter=0 >> /etc/sysctl.conf
echo net.ipv4.conf.default.rp_filter=0 >> /etc/sysctl.conf
echo net.bridge.bridge-nf-call-iptables=1 >> /etc/sysctl.conf
echo net.ipv4.ip_forward=1 >> /etc/sysctl.conf
sysctl -p

What is reverse path filtering?

Reverse path filtering is a mechanism adopted by the Linux kernel, as well as most of the networking devices out there to check whether a receiving packet source address is routable.

So in other words, when a machine with reverse path filtering enabled recieves a packet, the machine will first check whether the source of the recived packet is reachable through the interface it came in.

If it is routable through the interface which it came, then the machine will accept the packet
If it is not routable through the interface, which it came, then the machine will drop that packet.


Other References:
https://bugs.launchpad.net/fuel/+bug/1471647
https://bugs.launchpad.net/fuel/+bug/1539586