jwbensley's questions

What exactly are the meanings of the following settings when manually configuring network interfaces on CentOS (I'm trying to configure an IPv6 only host)?

I can guess what some of them mean by their name but I don't know what all of them mean (and some of my guesses could be wrong); is there any official documentation for these configuration options, I can't find any?

From /etc/sysconfig/network-scripts/ifcfg-eth0

IPV6INIT=yes          # I assume this just enables IPv6 networking on this interface?
IPV6_AUTOCONF=no      # Does this disable SLAAC?
IPV6_AUTOTUNNEL=no    # ?
IPV6_FAILURE_FATAL=no # Something to do with IPv6 not working not being an issue?
IPV6_DEFROUTE=yes     # ?
IPV6_PEERDNS=yes      # Does this mean we use the default gateway for DNS queries?
IPV6_PEERROUTES=yes   # Same as above?
IPV6ADDR=1::2/64      # This is obviously the IPv6 address and subnet mask
IPV6_DEFAULTGW=1::1   # This is obviously the default gateway

Also for IPv4 the settings DNS1 and DNS2 exist, do the equivilent ettings IPV6_DNS1 and IPV6_DNS2 exist?

I am currently building a software RAID under Linux using mdadm utility and I've read a few articles that describe how to increase the stripe_cache_size value for that RAID and how to calculate an appropriate value for stripe_cache_size.

I have increased mine to 16384 and my current sync rate on a new RAID5 in /proc/mdstat has jumped from from 71065K/sec to 143690K/sec (doubled!) which is good news. I also see the matching and expected increase in RAM usage however I can't find any documentation on what this setting does and how it works.

It seems to be a cache of some sort for the RAID that exists in RAM. That is all I can tell from its name and the effects seen by changing it. Is there any official "Linux" documentation for this setting and description of it?

There are many discussions and websites that explain the process of setting up a Linux software RAID with mdadm with the chunk size of a new RAID as 128kBs or 512Kbs. Serverfault is no exception.

I am now building a new media NAS and I see no good reason why I don't use a chunk size of 4kB though. Each of the four physical disks in the 'RAID-5-to-be' have 4KB sectors. Surely a 4kB chuck size makes the most sense mapping a 1:1 relationship from RAID volume down to disk sectors? Then atop that create the file system (which will be EXT4) with a 4kB block size?

How does a 128kB (for example) chunk size become more beneficial when disks only have 4kB sectors?

After reading on the Pound website that SSL does not support virtual host names, this in fact seems quite strait forward and obvious to me now. I have a pound load-balancer I wish to terminate SSL connections with. If a web server behind it runs multiple sites based on host name, can I configure pound to use multiple SSL certificates for these different hosts.

The only way to do this I can think of, is to assign another IP to the load balancer for each site that wants to use SSL, and configure a certificate for that IP & port combination for a specific site.

Is any one doing this, would this work?

UPDATE

Ideally, if anyone can share a config example that would be the best solution so I can read about it. Thank you.

I would like to build a Linux load balancer with SSL off load with sticky sessions. I would like to do this with either Pound or Pound and HAProxy. I haven't done this before, and I have been wanting to learn about both HAProxy and Pound for a while now. Finally I have a small use case for it to use as my excuse to get my hands dirty.

The site is a forum with a peak of ~4Mbps of throughput, which is a lot of posts and reads I think! So I don't want a high throughput device, I'm more concerned about concurrent users.

I have the following queries though;

1. Where does the bulk of the work load exist on the load balancer, is it the CPU decoding the SSL traffic, or RAM caching sessions for sticky sessions?

2. Following on from query 1, I have a spare server I would like to use, but how can I relate server hardware specification required against web application performance required?

   I have a small 1u server (PowerEdge 1850), with 2x76GB 10k Ultra 320 SCSI drives in RAID1, 2x 3Ghz single core Xeons (800Mhz Bus with 2MB L2 cache), and 6x 1GB sticks of PC2-3200 400Mhz RAM. I would like to use this, but I have no experience with HAProxy and Pound so I can't say if this will be apt or not. I would assume 6GBs of RAM is way too much looking at hardware load-balancer specs. What do others think about the CPU and HDDs?

This isn't a shopping thread, so don't post server models that would be suitable. Instead, what I would like if this isn't up to the task, is to go back to query (1) so I can build something else that will be sufficient. I have lots of experience with server deployments, but not these two packages.

Thank you.

I understand the concept of FCoE. I have looked at the Wikipedia page and looking at the layer 2 frame diagram, it looks like FCoE really should "just work" on any Ethernet switch, but is this really the case?

If so, what do switches like Cisco's Nexus 5k or 6120P offer that normal switches don't (in specific relation to FCoE)?

I am just using those two switches as examples. On the Nexus 5548UP page for example it says the following;

Unified ports that support traditional Ethernet, Fibre Channel (FC),and Fibre Channel over Ethernet (FCoE)

Well if FCoE runnins over regular Ethernet, that why does it support "Ethernet and Fibire Channel over Ethernet"?

This is why I am curious as to weather FCoE will run on any Ethernet switch and these switches just support "bonus" features, or if you do indeed require a specialist switch.

Thank you.

UPDATE

I have read on the Internet that FCoE will run on normal Ethernet devices but without DCB advanced features won't work, can anyone confirm if they have done this. Is it just "ill-advised" as opposed to flat out "won't work"?

I have a Windows Server 2008 x64 Standard virtual machine that runs on a machine with a hardware RAID controller, a Perc 6/i, which has a battery on-board.

Doing everything I can for additional performance, I think I should disable this. Is this very dangerous though?

My understand is that Battery Backed Write Caching gives a performance boost to the host OS, telling it the write was complete when they are still sitting in flash waiting to be written.

However, I can't see how it would be detrimental to performance, but is there a gain (even if marginal) to enabling it / disabling it?

P.s. There machine has a backup power.

Here is a screen shot for clarification:

I am trying to install Win2k8 Std as a KVM guest on Debian Squeeze.

As you can see from these screen shots;

1. No drives are detected (I have blanked out a 20GB image for testing) - screenshot1
2. I am using this driver CD: - screenshot2
3. I have signed the Win7 driver (I assume this was the most appropriate one?) - screenshot3
4. I can now see an unpartitioned drive - screenshot4
5. But I can't create a new partition on here, getting the error code 0x80070013 - screenshot5

I have had this error code before but only on a physical server. If I remember correctly it was complaining because the disks were partitioned as GPT (because it was a server that was being re-purposed) so repartitioning with an MS-DOS table fixed that. This is a blank disk image though. What is wrong here, and how can I correct this?

Thank you.

UPDATE

I have booted the VM with a Gparted-Live disk and formatted this volume with an MS-DOS partitioning scheme, and a single 20GB NTFS file system. Now when I boot the Win2k8 CD, load my drivers, I get a different error. As you can see at the bottom of screenshot6

"Windows cannot be installed on this hard drive space. Windows must be installed to a partition formatted as NTFS".

Clicking format produces the error (0x80004005) on the screen, so I think this is still a driver issue because Windows can see the drive but not interact with it properly. Is that insane thinking?

I am setting up a Debian Squeeze box as a KVM host. I want to add multiple interfaces to each KVM guest so I want them to be on different VLANs.

After reading about this, I believe the best method is to add multiple logical VLAN (sub)-interfaces to the physical NICs and then create a bridge adapter for each VLAN interace, and assign each bridge as a NIC for KVM guests. Does this make good sense, or madness?

Do I have to use bridged interfaces with KVM like this? Can't I just add eth1.xx and eth1.yy to my interfaces config below and then configure those directly as bridged KVM guest NICs? If so, how should this look in the interfaces config file below?

user@host:~$ cat /etc/network/interfaces 
# This file describes the network interfaces available on your system
# and how to activate them. For more information, see interfaces(5).

# The loopback network interface
auto lo
iface lo inet loopback

# Management Interface
auto eth0
iface eth0 inet static
address 172.22.0.31
netmask 255.255.255.0
gateway 172.22.0.1

# Interface for guest VMs
auto eth1

# Guest1 : Use VLAN 117
auto eth1.117
iface eth1.117 inet manual
# Set up br1 for guest 1, bridging with vlan 117
auto br1.117
iface br1.117 inet manual
bridge_ports eth1.117
bridge_stp off

user@host:~$ uname -a
Linux hostname 3.4.9 #1 SMP Wed Aug 22 19:08:46 BST 2012 x86_64 GNU/Linux

UPDATE

I would really like it if someone could clarify the config for me, as I have also seen the above configured with this syntax, so I don't see why one would be preferred over the other;

# Interface for guest VMs
auto eth1
allow-hotplug eth1
iface eth1 inet static

# Vlan 117 for guest 1
auto vlan 117
iface vlan111 inet static
vlan_raw_device eth1

# Guest 1 : NIC 1
auto br1.117
iface br1.117 inet manual
bridge_ports vlan117
bridge_stp off

ANSWER

I agree with dyasny on the topology, and am happily using the following config with is similar to Silopolis' example:

# Mangement Interafce
allow-hotplug eth0
# Guest Interface
allow-hotplug eth1

# Guest 1
allow-hotplug vlan116
# Guest 2
allow-hotplug vlan117

# Management
auto eth0
iface eth0 inet static
address 10.0.0.10
netmask 255.255.255.0
gateway 10.0.0.1

# Guest 1
auto vlan116
iface vlan116 inet manual
vlan_raw_device eth1

# Guest 2
auto vlan117
iface vlan117 inet manual
vlan_raw_device eth1

allow-hotplug br116
allow-hotplug br117

# Guest 1
auto br116
iface br116 inet manual
bridge_ports vlan116
bridge_stp off

# Guest 2
auto br117
iface br117 inet manual
bridge_ports vlan117
bridge_stp off

I have three KVM servers each with 1 guest VM, running directly on it's local storage, (so they are essentially getting a dedicated box worth of computing power each). In the event of a host failure I would like the guests replicated to at least one of the other hosts so I can spin it up there, until the failing host is fixed.

I am curious about KVM cloning. I can clone a VM live or when it's suspended/shutdown. Obivously suspended VMs will naturally be quicker to clone but these three VMs comprise three parts of a single solution, so I don't want to ever have any one of them shutdown.

How can I efficiently clone these VMs between servers?

I have had a couple of ideas, but are these insane or, is there a better method I have missed for my scenario?

1. Set up a DRDB partition between box 1 and 2 where VM 1 runs from, and so is replicated between box1 and box 2, repeat between box 2 & 3, and box 3 & 1 (This could be insane, I have never used DRDB only read about it)

2. Just use standard KVM CLI clone options to perform live clones (I'm dubious about this because I don't know how long it will take and what the performance impact will be during)

3. Run a copy of each VM on at least one other host, and have the guest on one host export it's data to the matching guest on another host where it can import that data, scripting this on the guest)

4. Some of other way? Ideas welcome!

Side Note

These servers have 4x15k SAS drives in a RAID 10 so they aren't rocketing fast, and as I mentioned, each VM runs from the host's local storage, no NAS or SAN etc. So that is why I am asking this question about guest replication. Also, this isn't about disaster recovery. Guests will be exporting their data to a NAS over a VPN, so I am looking at how I can have them quickly spun up in a host failure situation.

I updated nginx recently to version 1.2.2 and the following entry seems to be broken; presumably a change in syntax between versions?

location / {
                # First attempt to serve request as file, then
                # as directory, then fall back to index.html
                try_files $uri /index.html;
        }

When I browser to http://www.mysite.com/a-non-existent-url I get redirected to the a "500 Internal Server Error" error page. The following log entry is made;

2012/08/13 09:20:29 [error] 18457#0: *60 rewrite or internal redirection cycle 
while internally redirecting to "/index.html", client: 10.0.14.1, server: 
mysite.com, request: "GET /a-non-existent-url HTTP/1.1", host: "www.mysite.com"

This used to work as expected, I can't find the correct syntax for this version. How should this be now?

UPDATE Full config as request;

server {

    root /usr/share/nginx/mysite.com/public_html;
    index index.php index.html index.htm;
    server_name mysite.com www.mysite.com;
    access_log  /usr/share/nginx/mysite.com/logs/access_log;
    error_log   /usr/share/nginx/mysite.com/logs/error_log;

    location / {
        # First attempt to serve request as file, then
        # as directory, then fall back to index.html
        try_files $uri $uri/ /index.html;
    }

    #Added for awstats
    location ^~ /awstats-icon {
            alias /usr/share/awstats/icon/;
            access_log off;
    }

    #Added for awstats
    location ^~ /awstatscss {
            alias /usr/share/doc/awstats/examples/css/;
            access_log off;
    }

    #Added for awstats
    location ^~ /awstatsclasses {
            alias /usr/share/doc/awstats/examples/classes/;                                 
            access_log off;
    }

    #Added for awstats
    # Configure /cgi-bin/scripts to go through php-fastcgi
    location ~ ^/cgi-bin/.*\.(cgi|pl|py|rb) {
            gzip off;
            fastcgi_pass  127.0.0.1:9000;
            fastcgi_index cgi-bin.php;
            fastcgi_param SCRIPT_FILENAME    /etc/nginx/cgi-bin.php;
            fastcgi_param SCRIPT_NAME        /cgi-bin/cgi-bin.php;
            fastcgi_param X_SCRIPT_FILENAME  /usr/lib$fastcgi_script_name;
            fastcgi_param X_SCRIPT_NAME      $fastcgi_script_name;
            fastcgi_param QUERY_STRING       $query_string;
            fastcgi_param REQUEST_METHOD     $request_method;
            fastcgi_param CONTENT_TYPE       $content_type;
            fastcgi_param CONTENT_LENGTH     $content_length;
            fastcgi_param GATEWAY_INTERFACE  CGI/1.1;
            fastcgi_param SERVER_SOFTWARE    nginx;
            fastcgi_param REQUEST_URI        $request_uri;
            fastcgi_param DOCUMENT_URI       $document_uri;
            fastcgi_param DOCUMENT_ROOT      $document_root;
            fastcgi_param SERVER_PROTOCOL    $server_protocol;
            fastcgi_param REMOTE_ADDR        $remote_addr;
            fastcgi_param REMOTE_PORT        $remote_port;
            fastcgi_param SERVER_ADDR        $server_addr;
            fastcgi_param SERVER_PORT        $server_port;
            fastcgi_param SERVER_NAME        $server_name;
            fastcgi_param REMOTE_USER        $remote_user;
    }

    #Make sure all PHP is process by php-fpm
    location ~ \.php$ {
        fastcgi_pass 127.0.0.1:9000;
        fastcgi_index index.php;
        include fastcgi_params;
    }

    #rTorrent/wTorrent needs this
    #To loop back to the xml rpc service
        location /RPC2 {
                scgi_pass   127.0.0.1:5000;
                include     scgi_params;
                scgi_param    SCRIPT_NAME  /RPC2;
        }

}

2nd UPDATE

A debug log has beed posted here (http://pastebin.com/raw.php?i=PtLwvQhW). Its quite long so I have done this to avoid spamming this post.

I have various Dell servers (all 2950's) I would like to put some SSDs in. I have spoken to Dell on the phone and asked for a quote for 128GB SSDs for 2950's, they told me they don't sell them for this server.

These all have Perc 6/i's in them. Can I purchase any SSDs with a SATA II connection, of my liking, and connect this? Or will they be "blocked", or not work as expected because they don't have Dell firmware on them?

I have had (minor) issues before with HDDs that aren't Dell branded on Perc 5/i cards. They did work perfectly well (I sourced the exact same make and model drive as the failed original Dell ones, and bought them from a third party vendor), but at boot time the server would always display an warning about hard drive compatibilities (I assume was just being passed up to the OS by the RAID controller to scare me?).

I have a test box (PowerEdge 2950) with a Perc 6/i and 4x 15.5k SAS drives attached (with 512 byte block sizes). These are in a single RAID 5 virtual disk with a 64KB chunk size.

I am creating a single test partition that spans the whole drive. Should it be aligned to the 64KB chunk mark, or 512 byte block size? If the later, the partition could start at 2048 bytes into the single virtual disk, meaning it will begin at the 2nd free block on the first drive (I assume)?

Also, I will add another two drives and recreate the RAID virtual disk at a later date for more testing, should the partition then be created at 6x512 bytes, so from 3072 bytes?

I have read a couple of similar questions on this but I couldn't see from those, how the chunk size of the RAID volume might relate to partition alignment, on drive block size when using a single drive.

I have a Dell 2950 with a Perc 5/i controller inside it. I don't want to spark a debate about drive types and which is "better" etc etc. Just simply;

I would like to put in 2x Cheetah 15.5k SAS drives and mirror them for the OS. I would also like to put in 2x SSDs (exact model is still being debated) and mirror those for the data.

Can the Perc 5/i manage multiple RAID volumes like this, or do all drives have to contribute to the same single virtual disk/RAID volume, which you then split into logical volumes? If it can, is this desired set up possible on a Perc 5/i?

If I need a different RAID card (I don't want to start a shopping thread either), which Dell RAID card supports this requirement?

Also, is the throughput of a Perc 5/i going to limit the throughput of the SSDs?

Many thanks.

I have been reading about etckeeper, but I can't work out if you can use it on any other directories other than /etc.

There seems to be no reference to /etc in any configuration file (that I can find) so I assue this is hard coded and you can't chose directories to monitor. Is this true? I'd like to also have it manage /opt. Can this be done?

I have some Dell PowerEdge 2950's with 4x 15k, 150GB Cheetah SAS drives in them. They are going to be VM hosts, CentOS running ESXi with Windows Server 2k8 guests. Some guests will be hosting IIS servers, and others MSSQL servers.

I am trying to set the RAID virtual disks settings and can't decide which is more optimal given this situation;

Read Policy: Out of Read-Ahead, No-Read-Ahead and Adaptive Read-Ahead, the default is Read-Ahead.

I will be making large sequential writes initially, writing out blank images for virtual machine hard drives (lets say 30GBs from /dev/zero for example) so Read-Ahead seems good at first. But within the virtual machines reads could be random from anywhere within their file systems as they are IIS and MSSQL servers, so perhaps No-Read-Ahead is a better idea?

Now I think Adaptive Read-Ahead would be better then as a compromise but I don't know much about this option, how does it compare in performance to the others?

Write Policy: write-back caching, write-through caching, the default is  write-back caching.

The default of write-back caching is safer than write-through caching but at a performance expense. My thinking here is that in the event of power loss for example, it seems more likely in my head (this is why I need some clarification!) that damage will occur to a guest VM with write-back caching enabled, so I should favour write-through?

I have searched around and there is obviously no definitive answer, so I would like to find out what is best for my situation.

I see many articles on line stating that the overhead of Ethernet is 18 bytes (example: http://www.learnios.com/viewtopic.php?f=8&t=18887)

Looking at this wiki article (http://en.wikipedia.org/wiki/Ethernet_frame#Structure) 18 bytes is the following fields;

• Destination MAC
• Source MAC
• Ethertype
• CRC

Why do the preamble, start of frame delimiter and inter-frame gap not count towards this, are they not included in every frame? This would take the Ethernet overhead up to 26 bytes.

The original article I linked was with regards to calculating the required MTU for MPLS networks (encapsulating layer 2 traffic into MPLS frames). I want to understand why 18 bytes is being added on to the standard 1500 bytes and not 26 bytes (plus the MPLS label stacks and any do1q tags).

I can graph a device interface with SNMP tools, platforms like Cacti even make pretty graphs, but these are based on polling intervals (typically, every 5 minutes).

I can use the CLI;

r1>show int gi0/0
GigabitEthernet0/0 is up, line protocol is up 
Hardware is BCM1250 Internal MAC, address is 0011.2233.4455 (bia 0011.2233.4455)
Description: link 1
MTU 1540 bytes, BW 1000000 Kbit, DLY 10 usec, 
   reliability 255/255, txload 35/255, rxload 20/255
Encapsulation 802.1Q Virtual LAN, Vlan ID  1., loopback not set
Keepalive set (10 sec)
Full Duplex, 1000Mbps, media type is RJ45
output flow-control is XON, input flow-control is XON
ARP type: ARPA, ARP Timeout 04:00:00
Last input 00:00:00, output 00:00:00, output hang never
Last clearing of "show interface" counters never
Input queue: 0/75/3208/72484 (size/max/drops/flushes); Total output drops: 1373910311
Queueing strategy: Class-based queueing
Output queue: 0/1000/12 (size/max total/drops)
5 minute input rate 79658000 bits/sec, 19312 packets/sec
5 minute output rate 140174000 bits/sec, 21984 packets/sec

I can see from the 5 minute output rate I'm transmitting 140Mbps of traffic, but that's averaged over the last five minutes. So not right now, and no better than Cacti et all.

I could enter the interface command load-interval 30 to bring down the sampling rate to 30 seconds, at which point the txload and rxload values become more accurate.

Again though, if I need to find out which link is maxed out right now, I find it difficult to believe that for all the amazing things Cisco routers and switches can do, they can't tell me current tx/rx rate for an interface, right now.

I understand a sampling period maybe required for such a figure to be attainable, what's wrong with 1 second? Surely the CPU demand isn't that much, just counting the number of packets that pass every second and their size?

Has someone perhaps developed their own way of working this out? I see others haunted by the same conundrum?

UPDATE: I should have given better context to this, so I will try and do so now.

A typical need for me to know an interface throughput right now, is when a customer who has a 100Mbps port rings up and says "Hey, I'm download X from your mirror server but its only transferring at 20Mbps". I want to be able to see their switch pot throughput right now, whilst they're on the phone to me, to confirm this (non-tech customers often report incorrect values in my experience).

So, in this scenario I can confirm whether the customer port is already receiving 100Mbps so they only get 20Mbps because their port is at capacity, or if they are even transferring at anything like the speed they are claiming (whether it be above or below). Also, next I will want to see the throughput of the router that switch is terminated on for all those customers, this is another potential bottleneck. Also, I can check the switch port of the mirror server whilst the transfer is in progress.

I don't want to respond to the customer with "OK, well make sure you continue to download for another 5 minutes so I can wait for $NMS_OF_CHOICE to poll the interfaces", not an acceptable answer in the customers eyes. I can give many more scenarios, but essentialy, complaining customers is top priority :)

I have access to a CentOS 6 web server (a PHP/MySQL forum) running Apache 2.2.15 and MySQL 5.1.52. This box has dual oct-core Xeons.

What I find looking over the CPU graphs is that the first two cores (0 and 1) hit 90% usage whilst the other fourteen are only reaching 30-40% usage. Core 8, which I believe is the first core of the second processor, see's high usage (up to 90%) but then as mentioned, all others are low.

Apache fires up about 20 processes and I suspect in its default state, it can't understand, and make use of all the cores naturally. This should be the kernels job I think (can someone confirm?). However, with those cores maxing out, there is a noticable performance drop despite spare RAM and disk I/O.

Is there something I can tweak in Apache to "make it aware" of all the other cores, or perhaps a kernel peramater to assign certain process to certain cores? Can I exclude cores 0 and 1 for example, from being used for Apache and MySQL?

I have heard that "shutdown -h now" is not a graceful shut down command because it does something akin to "kill -9" for all processes, as opposed to "kill -15".

Is this true, and if so, what is a more graceful way of shutting down to ensure everything has time to clean up after its self?

P.s. Using CentOS, Ubuntu, and Debian mainly.