Troubleshooting PXE boot problems on a Dell server

At work, we ship "appliances"—servers that have a pre-installed operating system and software stack.  The general workflow to build one is:

1. The box is racked and plugged into the "build" network.
2. The box PXE boots to a special controller script.
3. Someone answers a few questions.
4. The script partitions and formats the disks and installs the OS and software.
5. The box is powered down, packaged, and shipped.

I recently spent days fighting with a new PXE boot "live OS", encountered error messages so lonely that few sites on the Internet reference them, debugged "initrd", and ultimately solved my problems by feeling really, really stupid.

I'll tell you the story of what happened in case you happen to try to walk down this path in the future, and I'll go into detail on everything as we go.  But first...

The Motivation

When you PXE boot, you get a "live OS", typically a read-only operating system with some tmpfs mounts for write operations.  This OS is often minimal, having just enough tools and packages to accomplish the goal of partitioning, formatting, and copying files.  Ours is done by mounting the root filesystem over NFS, so the smaller the OS, the better.

We had two problems that I wanted to solve:

1. The current live OS was hand-built on Ubuntu 12.04 with no documentation; and
2. Ubuntu 12.04's version of "lbzip2" is buggy and can't decompress some files.

My goal was to switch to Ubuntu 14.04 (which has a perfectly working version of "lbzip2") in a way that the live OS would be generated from a script (reliably and repeatably).

(If you're interested in building an Ubuntu filesystem from scratch that you can then use as a live OS, or make into an ISO image, or whatever, see debootstrap.)

The Execution

I created a few-dozen line script to bootstrap Ubuntu 14.04, install a kernel and some packages, and install the tools that it would need to run when it booted.  That part went well.

I copied the new OS to our NFS server, replacing the existing one.  That part went well.

All of our automated virtual-appliance builds continued to work (actually better and faster than before).  That part went well.

I met with the manufacturing team to get the formal sign-off that it would work with the physical Dell servers that we use.  This did not go well.

The Flaw In The Plan

Manufacturing booted up a box onto the build network but it crashed.

Here's what he reported:

Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000200

CPU: 4 PID: 1 Comm: init Not tainted 3.13.0-24-generic #47-Ubuntu

Hardware name:    /0JP31P, BIOS 2.5.2 01/28/2015

 ffff880802af0000 ffff8808041f5e48 ffffffff81715ac4 ffffffff81a4c480

 ffff8808041f5ec0 ffffffff8170ecc5 ffffffff00000010 ffff8808041f5ed0

 ffff8808041f5e70 ffffffff81f219e0 0000000000000200 ffff8808041f8398

Call Trace:

 [<ffffffff81715ac4>] dump_stack+0x45/0x56

 [<ffffffff8170ecc5>] panic+0xc8/0x1d7

 [<ffffffff8106a391>] do_exit+0xa41/0xa50

 [<ffffffff8109dd84>] ? vtime_account_user+0x54/0x60

 [<ffffffff8106a41f>] do_group_exit+0x3f/0xa0

 [<ffffffff8106a494>] SyS_exit_group+0x14/0x20

 [<ffffffff817266bf>] tracesys+0xe1/0xe6

------------[ cut here ]------------

WARNING: CPU: 4 PID: 1 at /build/buildd/linux-3.13.0/arch/x86/kernel/smp.c:124 native_smp_send_reschedule+0x5d/0x60()

Modules linked in: nfs lockd sunrpc fscache

Okay, a kernel panic.  I did upgrade the kernel from the old Ubuntu 12.04 version.  In fact, the old live OS had a hand-built Gentoo kernel for arcane reasons.  So maybe...

The Red Herring

Well, I did install the Ubuntu 14.04 ".deb" package for the Fusion I/O Drive's "iomemory-vsl" kernel module.  Some of our equipment have I/O Drives, so the live OS needs to be able to talk to them.

The VMs (which don't have I/O Drives) worked fine, and the physical hardware (which do) did not work.

I assumed that there was some mismatch between Fusion I/O's kernel module and the Ubuntu 14.04 kernel.  I compiled the driver (using Fusion I/O's guide) and re-made my live OS.

Same result: kernel panic.

In addition, it turned out that physical hardware that did not have an I/O Drive failed, as well.

The Hint

I took out my phone and recorded the boot process at 120 frames per second to see the text before the kernel panic.  The messages that I'd been able to see were not helpful, and when the kernel panics, I lose the ability to scroll up in the history.  And since the system never gets up all the way, there's no way to access the logs.

Here's what I saw:

systemd-udevd[329]: starting version 204

Begin: Loading essential drivers ... done

Begin: Running /scripts/init-premount ... done

Begin: Mounting root file system ... Begin: Running /scripts/nfs-top ... done

FS-Cache: Loaded

RPC: Registered named UNIX socket transport module.

RPC: Registered udp transport module.

RPC: Registered tcp transport module.

RPC: Registered tcp NFSv4.1 backchannel transport module.

FS-Cache: Netfs 'nfs' registered for caching

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

ipconfig: no devices to configure

/init: .: line 252: can't open '/run/net-*.conf'

Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000200

The short story here is that the "ipconfig" message repeating over and over means that the kernel can't find any network interfaces (at least other than "lo").

The long story is that when booting from network (NFS), the "initrd" init script is going to run some scripts to set up NFS.  One of these calls a function "configure_networking" (since NFS won't work without networking).  That function tries to figure out which device to use for networking.

For a box with multiple interfaces, this could be a bit tricky.  However, if you've toggled on bit b10 (0x2) in your PXE setup using the "IPAPPEND" directive, then you'll get an environment variable called "BOOTIF" that encodes the MAC address of the interface that originally PXE booted.  The "configure_networking" function then looks through the /sys/class/net/* files to see which interface has that MAC address.  From there, it will know the device and can set it up.

However, "ipconfig" claims that there are "no devices to configure".  Also, if you debug the "configure_networking" function, you'll see that there's only one file in /sys/class/net, and that's "lo".  So, the system has a loopback interface, but none of the Ethernet interfaces that I expected.

The Problem

Very simply put: the appropriate Ethernet driver (kernel module) was not present in the "initrd" file.

The "initrd" file (short for "initial RAM disk") contains some bootstrap scripts and kernel modules.  For example, for PXE booting, we get the kernel and the "initrd" file from TFTP (or HTTP) from the Ethernet card itself as part of the process.  Thus, it is up to the "initrd" file to get networking up and running so that it can mount the root filesystem over NFS.

I jumped through hoops trying to figure out how to get the "initrd" file to have enough (or the right) modules.  I eventually settled on bucking the trends set in the "Diskless Ubuntu" and "Diskless Debian" guides.

In my /etc/initramfs-tools/update-initramfs.conf, I set:

MODULES=most

instead of:

MODULES=netboot

The "MODULES" variable tells "update-initramfs" (and "mkinitramfs") which modules to include in the "initrd" file.  Obviously, less modules means less space.  So setting "MODULES" to "netboot" is a special setting that tries to pull in the bare minimum to get networking working.  I figured whatever; more modules is better than less modules, especially since the module for my Ethernet card isn't making it in there.

However, that still didn't help.

(Note: I haven't experimented with "MODULES=netboot" again; the "initrd" file that gets generated is under 20MB, and it takes under 1 second to transfer that on my network, so I'm not too interested in trimming that file down to the bare essentials at this point.)

My Ethernet card was some kind of Broadcom NetXtreme card, and the Internet seemed to think that the appropriate kernel module for it is "tg3".

After banging my head for way too long, I realized that "tg3" was not present in either the "initrd" file or the "/lib/modules/" directory.  I didn't realize that this was a problem because the Ubuntu 12.04 version of the system had compiled-in drivers (remember, I said that it was a Gentoo-built kernel); they were not built as modules (".ko" files).  So all of my grepping and comparing of files and directories seemed to tell me that a "tg3.ko" file was not necessary or relevant.

Once I realized that I needed the "tg3" module (and that it was compiled into the older kernel), I had to find out where to get it from.

The Solution

It turns out that there are two Linux kernel packages in Ubuntu:

1. linux-image; and
2. linux-image-extra.

The "linux-image-extra" package was key.  This package dropped in a lot of modules, including "tg3.ko".

So, in addition to installing "linux-image", I installed "linux-image-extra".  When I ran "update-initramfs", all of the modules got copied into the "initrd" file.  So when I booted the box, it successfully found my interfaces (for example, "eth2"), mounted the root NFS filesystem, and continued along its merry way.

FS-Cache: Netfs 'nfs' registered for caching

[...]

IPv6: ADDRCONF(NETDEV_UP): eth2: link is not ready

[...]

IP-Config: eth2 hardware address b0:83:XX:XX:XX:XX mtu 1500 DHCP RARP

IP-Config: no response after 2 secs - giving up

tg3 0000:01:00.0 eth2: Link is up at 1000 Mbps, full duplex

tg3 0000:01:00.0 eth2: Flow control is on for TX and on for RX

tg3 0000:01:00.0 eth2: EEE is disabled

IPv6: ADDRCONF(NETDEV_CHANGE): eth2: link becomes ready

IP-Config: eth2 hardware address b0:83:XX:XX:XX:XX mtu 1500 DHCP RARP

IP-Config: no response after 3 secs - giving up

IP-Config: eth2 hardware address b0:83:XX:XX:XX:XX mtu 1500 DHCP RARP

IP-Config: eth2 guessed broadcast address 10.XXX.XX.XXX

IP-Config: eth2 complete (dhcp from 192.168.XX.XX):

 address: 10.XXX.XX.XXX    broadcast: 10.XXX.XX.XXX    netmask: 255.255.252.0

 gateway: 10.XXX.XX.X      dns0     : 192.168.XX.XX    dns1   : 10.XXX.X.XX

 domain : XXX.XXXXXX.com

 rootserver: 10.XXX.XXX.XX rootpath:

 filename: pxelinux.0

Begin: Running /scripts/nfs-premount ... done

Begin: Running /scripts/nfs-bottom ... done

done

Begin: Running /scripts/init-bottom ... done

The Lesson

Well, what did I learn here?

1. "ipconfig: no devices to configure" means that the kernel doesn't think that you have any Ethernet devices (either there aren't any, or you're missing the driver for it).
2. "linux-image-extra" has all the drivers that you probably want (it certainly has "tg3").
3. When I see "Running /scripts/nfs-premount" messages, that file lives in the "initrd" file.  It is totally possible to take apart an "initrd" file (with "cpio"), make changes to the scripts, and then put it back together (again with "cpio").
4. If it seems impossible that your I/O Drive kernel module is causing networking-related kernel panics, then you're probably right and you have a networking issue somewhere.
5. Not being able to set up NFS in a network-booting setup will result in a "kernel panic".  This surprised me; I figured that I'd get to an emergency console or something.  Kernel panics usually point to deep, scary problems, not something as simple as not getting an IP address.

Google App Engine cron jobs run as no user

Google App Engine (GAE) is a great platform to use when developing just about any new application.  I use it for a couple personal projects, as well as for some fire department apps (dispatch and inventory management).

Today I'd like to talk about two things: users and cron jobs.

Users

One of the really convenient things about GAE is that it has built-in support for Google authentication (no surprise there).  This means that you can let GAE take care of your sign-in system (and trust me, handling the single-sign-on 3-way handshake isn't all that fun).  With GAE, you easily go one of two routes:

1. Certain paths in your web application are automatically required to have a logged-in user.  If someone tries to access a path without being logged in, GAE will redirect him to a sign-in window and then bring him back when he's done.
2. Server-side, you can ask GAE for log-in and log-out URLs, and you can direct a user to these at any time to have him log in or out.

I personally prefer the second route because there's no strange redirection and all of my endpoints behave as I expect them to.  For example, if I have a JSON REST API, a user (who has not signed in yet) can access the endpoint and be given a normal 400-level error (in JSON) instead of being redirected to a Google sign-in page.  My REST clients much prefer this.

To see if a user is signed in, a Java Servlet can call the "getCurrentUser()" method of the "UserService" instance.  If the result is null, then no one is logged in.  Otherwise, you get a couple of helpful methods to tell you about the user:

1. "getEmail()"; this returns the user's e-mail address.
2. "getUserId()"; this returns a unique numeric ID for the user.
3. "isUserAdmin()"; this returns whether or not the user is an administrator of the GAE application.

For my apps, user authentication is whitelist style.  I check the "@" portion of the e-mail address to see if the user is in one of the domains that I care about (I typically build apps internal to an organization that uses Google's mail system), and I check the administrator status to grant administrator powers to my admins.

If someone tries to access a sensitive API endpoint without being logged in appropriately, I'll send back a 400-level error stating that the user is not signed in with an appropriate account.

Pretty easy stuff.

UserService userService = UserServiceFactory.getUserService();

User user = userService.getCurrentUser();

if( user == null ) {

   // There is no user logged in.

} else {

   // The user is logged in.

   System.out.println( "User is logged in: " + userService.isUserLoggedIn() );

   System.out.println( "User is administrator: " + userService.isUserAdmin() );

   System.out.println( "User:" );

   System.out.println( "   Auth Domain: " + user.getAuthDomain() );

   System.out.println( "   E-mail: " + user.getEmail() );

   System.out.println( "   Federated ID: " + user.getFederatedIdentity() );

   System.out.println( "   Nickname: " + user.getNickname() );

   System.out.println( "   User ID: " + user.getUserId() );
}

Warning: you cannot call "userService.isUserAdmin()" if the user is not already logged in.  If you try to, then it will throw an exception.

Cron Jobs

Another thing that GAE can do is schedule cron jobs.  Basically, these are page requests that are scheduled like normal Linux "cron" jobs.  So if you need to have some task performed regularly, create an endpoint for it and schedule a job to access that endpoint.

Cron jobs act as if an administrator is making the request, so they can access all paths with "admin" requirements.  Howerver, you cannot check this using "userService.isUserAdmin()" because cron jobs do not run as any particular user.

To determine if a request is coming from the cron scheduler, you have to check for the "X-Appengine-Cron" header.  This header cannot be faked (except by admins); if you try to set this header, GAE will quietly remove it by the time that it gets to your Servlet.

To detect a cron job, you have to check for the header and make sure that its value is "true".

String cronHeader = request.getHeader("X-Appengine-Cron");

if( cronHeader != null && cronHeader.compareTo("true") == 0 ) {

   log.info( "Cron service is making this request." );

}

Ultimately, if I'm checking to see whether a user is allowed to access a particular section, I go through these steps:

1. (Assume no access at all.)
2. No user is logged in.
1. Is the "X-Appengine-Cron" header set to "true"?  If so, then allow administrative access.
3. A user is logged in.
1. Is the user a GAE admin of the application?  If so, then allow administrative access.
2. Is the user's e-mail domain in the whitelist of basic user access?  If so, then allow basic access.

Fixing the default Ubuntu snmpd configuration

SNMP is super helpful for performance and health monitoring of any production equipment.  It's lightweight, easy to understand, and very resilient when Bad Things happen to the network.  If you're not monitoring your production equipment with SNMP, then probably should look into that right away (we use SevOne NMS at work).

Getting "snmpd", the Linux SNMP daemon, up and running on Ubuntu is simply a matter of installing "snmpd":
sudo apt-get install snmpd;

Or is it?

Default configuration woes

Logging

By default, Ubuntu wants to log literally everything that "snmpd" does to syslog.  While I love the enthusiasm, this quickly leads to overflowing logs and the headache around them (plus it makes it impossible to find any event that's actually important).

How many times do you want to see messages like this in your logs?

Sep 11 16:48:23 your-server snmpd[19552]: Connection from UDP: [192.168.59.101]:49867->[10.129.11.219]
Sep 11 16:48:23 snmpd[19552]: last message repeated 199 times

The logging options are specified on the "snmpd" command line, and are thus configured in "/etc/default/snmpd".

The default logging settings are:

-Lsd

"-L" is for the logging options.  "s" is for syslog.  "d" is for the daemon facility.

What we want are these settings:

-LS 4 d

"-L" again is for logging options.  Capital "S" is for a priority-filtered syslog, with "4" being "warning-level or higher".  Again, "d" is for the daemon facility.

Port access

By default, Ubuntu locks down SNMP access to "localhost", so it's 100% useless from a monitoring perspective.  While I respect the security-mindedness displayed here, I need my boxes to actually respond to requests.

The access options are specified in the "snmpd.conf" file, which is located here: "/etc/snmp/snmpd.conf".

At the top of the file, there is a configuration item called "agentAddress".  By default, this limits requests to those originating locally.

agentAddress udp:127.0.0.1:161

There is usually a line following it that's commented out, and that's the one that we want.  Get rid of the line above and make sure that this one is enabled:

agentAddress udp:161,udp6:[::1]:161

This makes sure that any requests to port 161 (the standard SNMP port) will be allowed.

Permissions

Yes, yes, we should all be using SNMPv3's great user-based access-control mechanism, but for an internal-to-the-company server that can't be reached from the Internet, we can often afford to be lax.  And hey, I'm not stopping you from setting up SNMPv3 access control.  Go nuts.

Here, we're going to allow the community string of "public" to access everything about the box (but not make any changes at all).

The default configuration allows "public" to see some basic system information, but that's not good enough:

rocommunity public default -V systemonly

Get rid of that line and replace it with one that doesn't have the "systemonly" restriction:

rocommunity public

Restart "snmpd" and you'll be ready to respond to SNMP requests from your local management station.
sudo service snmpd restart;

Google App Engine and Google Authentication

One of the things that I love about the year 2015 is "cloud computing"; in particular, I love that I can hand Google App Engine a Java project and it will host it, handle redundancy, auto-scale it, provide me a database, and do just about everything else that I could ever want.

However, security is still a major concern, and there are some applications that I work on where "leaking" some private data onto the public Internet is bad news.

User authentication

There are lots of ways to authenticate users at this point.  Years ago, we did everything ourselves (remember those days?).  Each application had its own user database with varying degrees of security, and passwords were being stolen and sold all the time.  Now we have things like OAuth, where we can pass off user authentication to other systems (which we have to trust), so we don't have to store anything more than an e-mail address.  If the OAuth server says that that e-mail address is legit and logged in, then it's legit and logged in.  This saves us time and money, since who wants to build and maintain a user authentication layer, anyway?

Google App Engine provides a pretty easy and awesome way to lock down your application to "signed in" users.  Here, "signed in" users can be one of two things:

1. Any user on Earth with a Google account; or
2. Any user in your Google Apps domain.

If you have a Google Apps domain, then with a couple of tweaks to both the domain and the app, then Google will make sure that only those people in the domain can log in to the app.  Otherwise, if you'll have to check for a particular domain from your logged-in users in your REST API calls (your app is RESTful, right?).  No biggie, either way.

(In this article, I'll be talking about the Java version of the Google App Engine SDK, so any files or calls will be related to that.)

To set up your app to force everything to require a logged-in user, you just need to update "web.xml" and add a "security-constraint" (obviously, you can play a lot with this):

<security-constraint>

<web-resource-collection>

<web-resource-name>site</web-resource-name>

<url-pattern>/*</url-pattern>

</web-resource-collection>

<auth-constraint>

<role-name>*</role-name>

</auth-constraint>

</security-constraint>

The "url-pattern" of "/*" means "MATCH ALL THE URLS!", and the "role-name" of "*" means "all users signed in with a Google account".

I can't log out!

I recently put together an application for a group where I don't own the Google Apps domain, so I couldn't set up the sweet domain-level restriction at the app level.  Like I said, it's no big deal, so at the top of my API class was a check for "@myspecialdomain.org" in the user's e-mail address (remember, Google already verified that it was a Google account, so I just needed to make sure that it was in the right domain).

I asked the group to try out the application, and half loved it; the other half couldn't see any data.  It turned out those people logged in with their personal accounts (not their organization-specific accounts), and now were stuck in an application where they had no access to any data.

What I needed was a "switch accounts" button (like all of the other Google applications), or at least a way to invalidate the user's session so that they could log in again (and use their organization account, this time).  Google App Engine's built-in authentication system does some magic, so the usual Google authentication guides don't apply.  I tried all of the stuff that people online were talking about (invalidating sessions, deleting cookies, etc.), but none of it worked, and it seemed like a hack anyway.

It turns out that Google App Engine provides a "UserService" class help deal with this kind of problem.  In particular, you may request a login URL and a logout URL.  Since my whole app requires the user to be logged in, all I needed to do was log the user out, and he would be immediately redirected to the Google login/pick-account screen again.

I added a simple API call to return some information about the current user (so I could show the "Logged in as ..." message), and that call returned an additional property for the logout URL (remember, this URL is generated by Google App Engine and isn't trivial).  When my UI loads, it makes an AJAX call for the user information, grabs the logout URL, and provides a "Logout" link for any users that need to log out.

Problem solved.

Here's how to get the logout URL:

UserService userService = UserServiceFactory.getUserService();

String logoutUrl = userService.createLogoutURL("/");

The "/" argument to "createLogoutUrl" is where the user should be redirected once he logs out.  In this case, I'm sending him right back to the main screen of the app, where he will be asked to log in again.

Additional resources

1. For more information on "web.xml" within Google App Engine, see this document.
2. To learn how to set up authentication with Google App Engine, see this article.

Beware the AppScale firewall

If you haven't already looked into AppScale for your company's internal application needs, then you may want to spend some time looking into it.  In short, it's an open-source implementation of Google App Engine that can run on a "private" cloud.  Why is this cool?  Well, it lets me use all of the power and convenience of Google App Engine apps without having to put my app in the public cloud (high latency and billing), instead letting me use all the resources that I want in my company data center.

tl;dr: AppScale runs "iptables" on its own, so if you want to run an additional service (such as SNMP) on a node, then you'll have to configure AppScale to allow it.

My goal: SNMP monitoring

At work, I'm setting up an AppScale cluster to serve some internal applications.  The first thing that any server needs to do, once online, is provide performance statistics (via SNMP) to our performance management tool (in our case, we use SevOne NMS).

Typically, this is the world's easiest task:

1. Install "snmpd" (apt-get install snmpd).
2. Allow "snmpd" to respond to remote requests (duh).
3. Fix Ubuntu's terribly verbose SNMP logging defaults.

Unfortunately, I fought with this for an hour because, no matter what I did, "snmpd" would not respond to any requests from my management tool, and nothing in the logs said why.  For reasons not perfectly clear to me, AppScale (for Ubuntu) runs on Ubuntu 12.04, so I thought that maybe there was some ancient security measure in place that I had forgotten about over the years.

I eventually stumbled on "iptables" as a culprit (it's never first on my list, but probably should be).  I ran "iptables -L -n" to list the current "iptables" rules, and sure enough, the system had some:

Chain INPUT (policy ACCEPT)
target     prot opt source               destination        
ACCEPT     all  --  0.0.0.0/0            0.0.0.0/0          
ACCEPT     all  --  0.0.0.0/0            0.0.0.0/0            state RELATED,ESTABLISHED
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:22
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:80
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:443
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:1080
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:1443
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:2812
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:5222
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:5555
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:6106
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpts:8080:8100
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpts:4380:4400
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:17443
ACCEPT     tcp  --  0.0.0.0/0            0.0.0.0/0            tcp dpt:4343
ACCEPT     all  --  10.129.11.219        0.0.0.0/0          
ACCEPT     all  --  10.129.11.221        0.0.0.0/0          
DROP       all  --  0.0.0.0/0            0.0.0.0/0          

Chain FORWARD (policy ACCEPT)
target     prot opt source               destination        

Chain OUTPUT (policy ACCEPT)
target     prot opt source               destination

However, no amount of "iptables" magic would allow me to get the system to respond to SNMP requests.  I'd add a rule, and it might respond for a few seconds, but after that, my SNMP requests would time out again.  My rule?  Gone.

The AppScale firewall

It turns out that AppScale maintains the "iptables" setup for the box, and any change that you make will quickly be reverted by it.  This doesn't, in principle, bother me except that it's not really documented anywhere.  The only real mention of it is the Performance Tuning document, and even then, it's just a quick mention in order to get HAProxy stats from the box.

The AppScale firewall configuration lives in "appscale/firewall.conf" (the default installation guide had me put the "appscale" directory in "/root", so the file was located in "/root/appscale/firewall.conf" for me).  Once I saw what was going on, it was simply a matter of making a quick change to the file and waiting a few seconds (AppScale periodically re-reads the file and makes any changes live).

To tell AppScale to allow SNMP requests, I simply had to add the following line after the other "iptables -A" lines:

iptables -A INPUT -p udp -m udp --dport 161 -j ACCEPT # SNMP

Problem solved.

Additional resources

1. The current default version of AppScale's "firewall.conf" can be found here.

Chrome/Chromium, Roboto, and the horrible text nightmare

I began doing Android development a year or so ago, and the first thing that I noticed was that the official Google Android documentation looked absolutely horrible in Chromium (and Chrome too).  I tried updating fonts, disabling fonts, and installing new fonts, and none of that helped.  On other people's computers, the text looked fine.  In Firefox on my own computer, the text looked fine  On my computer, in Chromium: total crap.

One of the directions that I researched was around the Roboto font (which apparently was released with Android Ice Cream Sandwich).  However, it turns out that the font itself has nothing to do with the problem.

Today, I finally figured it out (and fixed it!), and my Internet (and Android development) experience has been much, much better.

First, let me tell you that I've been running Kubuntu (the KDE Ubuntu variety) this whole time.  So, this covers Kubuntu 11.10, Kubuntu 12.04, Kubuntu 12.10, Kubuntu 13.04, Kubuntu and 13.10.  I had the problem with all of them.

Here's what it looks like:

To make a very long story short, I had to enable anti-aliasing for fonts for the entire system.  I typically set all of my graphics settings to the lowest possible levels in order to not have to see stupid animations or other things that slow down my experience just to be on par with Windows, and it looks like font anti-aliasing is one of the settings that got turned off.

So, a quick tweak to the drop down box here:

And everything now looks nice and pretty.  Here's the exact same page from before, this time with system-wide anti-aliasing enabled:

Problem solved.

Beware the void* trap

I'd like to take a little bit of time today to share a problem that took me numerous days to track down and solve because it was so obscure and stealthy.  At work, we use ZeroMQ to handle any inter-process communication, and our primary language is C++.  I had begun a fairly intense project to remove ZeroMQ from the intra-process communications of a particular daemon, leaving it only for when we need to cross process boundaries.  Basically, once the messages that we want get into our daemon, there is no reason to pass them off to threads and such by serializing them to byte arrays (since that what a ZeroMQ message is) when they could be added (as objects) the lists, passed around to thread pools, etc.

As I was nearing the completion of the project, I found that all outbound communication from one of the sections of code seemed to be lost.  This was strange because the inbound communication was handled properly.  I had refactored both directions, so it was quite possible that I had messed something up.  However, no matter how many times I checked the logic, no matter how many log statements I made at each line, nothing looked amiss.  I struggled for days trying to understand why the "send()" calls seemed to be going nowhere, and the answer shocked me.

Here is a simplistic version of the code that I had.  Basically, this code receives (from somewhere else in the program) a list of messages to send to a ZeroMQ socket as a single atomic message.  This is accomplished by adding the "ZMQ_SNDMORE" ("send more") flag to the call.
void sendMessageToEndpoint1( std::list<zmq::message_t*>& messages ) {
   while( messages.size() > 0 ) {
      //! This is the ZeroMQ message that we're going to send.
      //! It has been prepared for us elsewhere.
      zmq::message_t* message = messages.front();
      // Remove the message from the list.
      // The size of this list is now the number of remaining messages to send.
      message.pop_front();
      
      //! This contains the flags for the "send" operation.  The only flag that
      //! we're actually going to set is whether or not we have more messages
      //! coming, and that's for every message except the last one.
      int flags = messages.size() > 0 ? ZMQ_SNDMORE : 0;
      // Send the message to the endpoint.
      // Note that "endpoint1" is of type "zmq::socket_t*".
      endpoint1->send( message, flags );
      
      // We can now delete the message.
      delete message;
   }
}

I promise you that I had logged something before and after every statement, and everything was exactly as I expected it to be.  No exceptions were thrown.  There were no compiler warnings.  But no client on the other side of "endpoint1" ever got any of the messages that were being sent.  This drove me crazy.

The answer is that I was passing the wrong thing to "zmq::socket_t::send()".  Unlike the "recv()" ("receive") call, which takes a pointer to a "zmq::message_t", the "send()" call merely takes a reference to a "zmq::message_t".  Clearly this is a type error, so the compiler should have caught it.  But it didn't.

Here's the signature of the "send()" function.  Basically, it sends a message with some optional flags.
bool send( zmq::message_t& message, int flags = 0 );

I was sending it a "zmq::message_t*" and "int", so the compiler should have reported an error, since the type of the first argument was incorrect.  However, no error (or warning) was printed, and it compiled fine.  Even stranger, nothing bad happened when I called "send()".  Nothing good happened, either, but the code ran with the only strange symptom being that my "send()" calls seemed to do nothing.  The client on the other side of the ZeroMQ socket simply never received the message.

So, what's up with that?

It turns out that there is another "send()" function, one that takes three parameters.  It sends an arbitrary number of bytes with some optional flags.
size_t send( void* buffer, size_t length, int flags = 0 );

And there's the rub.

We've already established that the first "send()" function shouldn't work.  But here's a second "send()" function that does meet our signature.  As for the first parameter, a "zmq::message_t*" will be implicitly cast to "void*" in C++.  As for the second parameter, "int" will be implicitly cast to "size_t", which is just an unsigned integral type.  As for the third parameter, it is not specified, so it'll be set to zero.

This second "send()" is clearly not what I wanted to use, but the compiler doesn't know that I thought that the function required a pointer to a message, not a reference to it.  Since "ZMQ_SNDMORE" is defined to be the number 2, this call to "send()" only attempts to transmit two bytes.  And because a "zmq::message_t" is certainly larger than two bytes (it is actually at least 32 bytes), the data to copy, from the second "send()" function's perspective, is always present.  This means that in addition to not getting any warnings or errors, I am also guaranteed to have this code never crash, since it will always send the first two bytes of the "zmq::message_t" structure.

Naturally, the fix was to send the dereferenced version of the message, and everything worked fine after that.  The moral of the story here is to watch out for implicit "void*" conversion.  And if you are making a library that accepts a byte buffer for reading/writing purposes, please set the type of that buffer to some byte-oriented type, such as "char*" or "uint8_t*".  These would require explicit casts, thus preventing accidental use as in my case.