chen's questions

I wanted to have an HA forward proxy solution using Squid, and I am trying to use the Squid servers behind ELB solution on page 41.

However, my forward proxy service is a service meant to forwarding traffic to an internal network (back to my corp network), i.e., there is no Internet connection for my Squid servers. Therefore, health check about a well-know URL such as google.com does not work.

With the solution describe on page 41 above what is the best way for ELB to do health checking toward the Squids servers? Which ping path I should be using in this scenario?

If I have an application looks like the following:

Client <--> ELB <--> EC2

Would the E2E latency possibly be lower if I set up ELB as TCP passthrough mode than I make it as HTTP listener?

The reason I guess TCP passthrough mode may make my E2E latency better (lower) is because ELB in this scenario almost does not cause any extra hop cost than the following scenario:

Client <--> EC2

Is my understanding correct? Please walk me through if not.

I saw AWS architecture that IGW is placed in front of ELB (i.e., see page 8 in http://www.slideshare.net/AmazonWebServices/high-availability-application-architectures-in-amazon-vpc-arc202-aws-reinvent-2013); and I also see ELB is placed directly to Internet connection (i.e., see page 14/15 at http://www.slideshare.net/AmazonWebServices/stg305)

What's the use cases for either architecture? or what's the pros and cons for either architecture design?

I have two servers (attached disks have been set up as hardware RAID1 for disk device level failure resilience).

Here is the setup in my mind:

1) Install ESXi on each of the physical server, M1, M2;

2) Start one VM on each of the ESXi virtualized physical server V1, V2;

3) Install the DRDB drivers within V1 and V2.

Essentially, this is a "virtualizing machine running DRBD in the VM's instead of bare metal hardware" idea.

My question is whether the above setup can achieve the same "networked RAID1" goal that DRDB can achieve in the bare-metal physical machines (http://www.drbd.org/).

Thanks.

[EDIT] I found this (DRBD experimentation and virtualization) is a similar question, but the answer does not seem to be firmative enough for me to follow.

[EDIT 2] One thing I need to understand is whether the DRBD driver inside the VM got the "Ok, data flushed to local disk" echo means "the data has been flushed to underlining physical disk". Because I want to consider corner situations like 'split brain problems'

I like Ganglia for its Gmetric function, and I like its multi-layer capability.

But Ganglia does not have healthiness check, alerting and etc. for the server monitor functionality. So it would be great to bring this two species together.

Sure, we can install Ganglia, and then install ZenOSS. But is there a plugin or something that can smoothly integrate them together? At least integration at the presentation level.

Thanks

I am experiencing the above problem and I have no idea what caused this. Can someone help me figure this out?

1) I have already had my first disk /dev/sda working properly, now I am trying to create another FS on the 2nd disk:

$fdisk /dev/sdb [I pressed n, p, w under fdisk]

2) I took a look at the result of fdisk

$fdisk -l

Disk /dev/sda: 1999.0 GB, 1998998994944 bytes
255 heads, 63 sectors/track, 243031 cylinders, total 3904294912 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x000a6eb5

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *        2048  1953124351   976561152   83  Linux
/dev/sda2      1953126398  2070310911    58592257    5  Extended
/dev/sda5      1953126400  2070310911    58592256   82  Linux swap / Solaris

Disk /dev/sdb: 1999.0 GB, 1998998994944 bytes
31 heads, 7 sectors/track, 17992142 cylinders, total 3904294912 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x49802e7d

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048  3904294911  1952146432   83  Linux

3) I make the following filesystem

$ mkfs.xfs -i size=1024 /dev/sdb

4) Now I see the problem...

$tune2fs -l /dev/sdb1 |grep -i inode
tune2fs: Bad magic number in super-block while trying to open /dev/sdb1
Couldn't find valid filesystem superblock.

I read the VMDirectPath materials, and they say that it helps the performance on the 10Gbs card case. Sure, if multiple VMs directly access the physical 10GbE card, the VMM bottleneck is avoided with VMDirectPath.

Here I have two questions:

1. If I have 3 1GbE cards and want to utilize the VMDirectPath benefit, should I first bind the 3 cards? Or should I somehow bind a VM to a physical NIC with the help of VMDirectPath? Is it possible?

2. If I want to configure my disks into passthrough mode in ESXi, can the performance get boosted?

I have a linux box with two IP addresses (say IP1, and IP2).

I have a server-process listening to a specific port (say 60000). Now, I want to run two copies of the server-process on the above linux box. How can I set up my system such that I can have process1 handle packets to IP1:60000 and have process2 handle packets to IP2:60000?

Or can I?