Category: Network & Server (網絡及服務器)

Riverbed Stingray (Zeus) Traffic Manager Load Balancer vAppliance

By admin, February 11, 2013 21:32

What can I say? Riverbed Stingray Traffic Manager (previously know as Zeus Traffic Manager)  is probably one of the best Load Balancers I’ve tested on the market, now it’s also available as VMware OVF format.

From initial VM OVF deployment to the actual implementation of the first load balanced web sites was less than 30 minutes. Simple concept and solid performance during stress tested (concurrent over 10,000 users per seconds).

I particularly like the admin GUI, simple and intuitive, everything is self explanatory and no non-sense.

What’s best, it also comes with an integrated Application Firewall as well as Traffic Manager Cluster Mode which you can add one or more Traffic Managers to the cluster to make it HA, wow, this idea is brilliant!

I’ve even gone as far as adding the 3rd TM (on a different subnets) into the cluster, it worked somehow, but failed when creating a Traffic IP as the Traffic IP must be seen on all subnet, I think this can be easily solved using a router between the two different subnets thought. Alternatively, I think GSLB is my next option on the list, but don’t have time now, will try later.

The most magical feature is Aptimizer which transparently optimize your web page (ie, compressed, hence reduced the time for loading) without you rewrite any of the coding.

Stingray Aptimizer is what was formerly known as Website Accelerator, or WAX. It was created by New Zealand software developer Aptimize to rejigger and accelerate web pages running on IIS or Apache web servers as well as pages stored on content delivery networks from Akamai Technologies.

Aptimizer analyzes how web pages load and reorganizes the content so a web browser doesn’t have to make so many roundtrips back to the web server to load a page. Because there are dozens of elements on a typical page, reconfiguring the web pages on the fly and storing the more efficient web page in cache on the web server can reduce page load times by a factor of four. The beauty is that this optimization does not change the web applications one bit, so you don’t have to modify your code.

The only complain is probably the cost, which is prohibitive to implement for the average SMBs, but since Riverbed is really targeting enterprise market, so I guess they don’t really care about the little ones after all. :)

Oh…there is a free and fully functional with limited features (10 request per second) Developer edition, don’t forget to try it out.

tm

Equallogic: RAID 5 and No-Spare Configurations Not Recommended

By admin, February 6, 2013 11:29

This is the latest recommendation from Dell Equallogic, found in Firmware 6.0 Release Note:

Beginning with this release, the Group Manager GUI no longer includes the option for configuring a group member to use RAID 5 for its RAID policy. RAID 5 carries higher risks of encountering an uncorrectable drive error during a rebuild, and therefore does not offer optimal data protection. Consequently, Dell recommends against using RAID 5 for any business-critical data.

RAID 5 may still be required for certain applications, depending on performance and data availability require-ments. To allow for these scenarios, you may still use the CLI to configure a group member to use RAID 5.

For a complete discussion of RAID policies on PS Series systems, review the Dell Technical Report titled PS Series Storage Arrays: Choosing a Member RAID Policy, which can be downloaded from either of the following locations:

• www.equallogic.com/resourcecenter/documentcenter.aspx

• en.community.dell.com/techcenter/storage/w/wiki/equallogic-tech-reports.aspx

In addition, Dell recommends against using RAID configurations that do not use spare drives. You should convert all group member that are using a no-spares RAID policy to a policy that uses spare drives.

Known Issues and Limitations

The following restrictions and known issues apply to this version of the PS Series Firmware. For information about known issues and restrictions from other releases, see the Release Notes for those versions. For issues about Dell EqualLogic FS Series Appliances, refer to the Dell EqualLogic FS Series Appliances Release Notes. For issues and limitations pertaining to host operating systems and iSCSI initiators, refer to the iSCSI Initiator and Operating System Considerations document.

RAID Conversion From No-Spares To Spares Does Not Work
RAID conversion from a no-spares policy to a spares policy appears to work, but it actually converts to no-spares, resulting in no change. (Funny :) )

SAN HQ 2.5, HIT Microsoft 4.5 and HIT VMware 3.5

By admin, February 6, 2013 10:36

I am pretty happy with SANHQ 2.2, HIT MS 4 and HIT VMware 3.1, and there aren’t much new features I need, so I choose to delay the upgrade for the time being.

Furthermore, there are two interesting videos presented by Dell EQL User Group in Taiwan (in Chinese), the following new features were intensively mentioned as well. (Part 1 and Part 2)

SAN HQ 2.5 Announcement

Dell announces Host Integration Tools for Microsoft 4.5

Dell Announces Host Integration Tools for VMware 3.5

Note: HIT for VMware v3.1.1, and earlier, is not compatible with Equallogic Firmware Version, 6.0, of the PS Series Firmware. A later version must be installed prior to upgrading to Version 6.0, for compatibility.

Windows Server 2012 Hyper-V Component Architecture Poster

By admin, January 29, 2013 11:50

This huge PDF poster shows some of the advanced features that VMware doesn’t have, even I am not a Microsoft visualization guy, but it’s good to know what competitor is up to.

1

Some Thoughts about Network/Storage Infrastructure Design

By admin, January 26, 2013 19:28

“Keep It SIMPLE!”

I was once told by a data center veteran about 10 years ago.

This has been deeply embedded in my network/storage/security design methodology ever since and it is still holding the truth even today. It saved me from many pitfalls and allowed me to quickly track down where exactly the problems were.

It is especially true when your equipments (server/switch/storage) are growing at a fast pace and “Keep It SIMPLE!” is the only tool to save your life after all.

Ops…After So Many Years, Now HKIX Announced New Charging Model

By admin, January 22, 2013 10:27

Important HKIX Announcement on New Charging Model

HKIX was established in April 1995, using spare resources of The Chinese University of Hong Kong.  At the onset, HKIX could be offered as a free-of-charge service to the early participants with moderate network traffic.  But over the years, with the growth in the demand for Internet traffic exchange and e-business, we have now grown into a critical Internet infrastructure covering 190+ autonomous systems and exchanging 250+ Gbps of data at peak time.  To support such a huge network, the configuration of HKIX has also evolved from a primitive coaxial cable to Ethernet switches and to the now top-end equipment and technologies.

Obviously, it is not possible to leverage the network economy of the whole Hong Kong with virtually no cost.  In 2005, we introduced the penalty-based cost-recovering model with a hope to curb abuse and to recover most of the cost.  After eight years, we find that this extremely conservative model is not sustainable in front of the exponential growth in network traffic volume and growth in the demand for new connections.  Together with the higher expectation of all the participants in the service, it calls for a continuous upgrade of the HKIX infrastructure in capacity and resilience.  This requires substantial periodic capital investment and operating expenses.  It is unfair and impossible for the university to subsidize the operation indefinitely.  In fact, we have to do a major upgrade in 2013 to cope with the ever-increasing demand and to improve our resilience further with site resilience at the core.

In order to make the HKIX operation truly self-sustained in long term and to prepare for the continuous growth in the future while we could keep HKIX as a not-for-profit service, we have no choice but to implement a full-cost-recovery model, i.e. simple port charge model.  This charging mechanism is actually the industry norm, which is fair and consistent for all participants.

Simple port charge model will be implemented starting 01-JAN-2013 for all new participants and new connection requests. For existing participants with no change of connections, the existing terms and conditions will not change until 30-JUN-2013.  To further ease the transition, we shall still provide to each participant up to 2 free FE/GE ports in total (counting all ports at all HKIX sites) until 30-JUN-2014.  After 30-JUN-2014, all ports will be chargeable.

The new port charges (applied to all HKIX sites) will be:

FE/GE: US$120/port/month (no one-time connection charge) 10GE: US$1,000/port/month (plus one-time non-refundable connection charge: LR – US$2,300; ER – US$7,000)

(Exchange Rate: US$1=HK$7.8)

We strongly believe the charges laid down are on the low side when compared to other parts of the world.  Nevertheless, we shall review the charges periodically and uphold our principles of fairness and consistency.

If you have any queries regarding this new arrangement, please feel free to contact our core team at hkix-core at itsc.cuhk.edu.hk.

We thank you for your understanding and wish you’ll have a Merry Christmas and a Happy New Year.

HKIX Team

Userful Command Line to Find Out the BAD Processes/Services

By admin, December 28, 2012 12:30

It’s a very common scenario that a server/PC maintained by careless keeper got hacked, which often leads to high CPU usage due to Trojan process running in background.

The followings two are the most useful tools to quickly determine which are the bad nuts.

netstat -nop tcp

- This is very useful to find out if a Trojan service has opened an incoming port, often the status shows “ESTABLISHED”

- netstat can be also used to find out the DDOS IP as those status often ends with “LAST_ACK”

- From the output PID, then you can determine which process from the task manager, but what if it’s a svhost.exe? Problem is you still can’t tell exactly which service as there are several svhost.exe running most of the time, so keep reading.

tasklist /svc |find “Process_ID”

Bingo! This cmd does exactly the trick to find out which is the bad nuts, Process_ID or PID is the one you got in netstat.

Is that so simple? I am afraid not, I’ve seen recently Trojan/DDOS tools have becoming more sophisticated and difficult to catch as they embedded themselves as web scripts utilizing UDP ports as the protocol, either acts as a bot waiting to be activated via URL strings and then send UDP DDOS to the target IP (ie attack others) or exhausts the entire server CPU by issuing UDP commands to fill up Events Logger (ie, attacked the server itself).

What you will find is w3wp.exe (or IIS Service) is using a lot of CPU as well as System (which is the disk activity) being very busy, so you got to use other methods to track down those bad scripts.

Finally, a single svchost.exe can contain many individual services, so the only way to find out which service is responsible is to configure them to run in its separate svchost.

For example, lets start with wuauserv (Windows Update Service) which is often the cause of high CPU, open command prompt and run the following commands:

net stop wuauserv

sc config wuauserv type= own

net start wuauserv

This will isolate wuauserv in its own svchost and then you can check the effect on CPU and identify the PID with tasklist /svc, you are already familiar with the process.

Taobao (淘寶) Security Hole? PHP Code Exposed!

By admin, December 13, 2012 18:53

This is so strange when I typed in “18 Mulsanne” in Taobao today and suddenly the result page turns into raw PHP code, it must be a serious security hole somewhere, probably it’s server got hacked. However, consequence searches all return normal result, just wired!

<?php

include "common.inc.php";
include "{$ROOT}/common/base32.inc.php";
$location_url = "/search?";
if ((isset($_SERVER["SCRIPT_NAME"]) && $_SERVER["SCRIPT_NAME"] ===
"/browse/search_auction.htm") || isset($_GET["taoke_from"])
&& $_GET["taoke_from"] === "search_auction") {
    foreach ($_GET as $k => $v) {
        if (!empty($v) && array_key_exists($k, $PARAMS)) {
            if (in_array($k, array('q', 'loc'))) {
                if (isset($_GET["_input_charset"]) && $_GET["_input_charset"]
=== 'utf-8') {
                    $v = urlencode(iconv('UTF-8', 'GBK', $v));
                } else {
                    $v = urlencode($v);
                }
            }
            $location_url .= $k . "=" . trim($v) . "&";
        }
    }
    $location_url = rtrim($location_url, '&');
    header("Location:$location_url");
    exit();
}
if (($_GET["cat"] <> "
...

First Equallogic Disk Failure in Two Years

By admin, December 12, 2012 18:24

Just received three email alerts simultaneously from SAN HQ, Dell OME and EQL Group Manager all saying slot 5 in one of the Equallogic members has failed, the last disk (slot 15) kicked in and the raid set is reconstructing. Called local Dell ProSupport, parts is being arranged and will be delivered to data center within 2-3 hours. (Update, raid reconstruction took about 3 1/2 hours to complete)

It was quite a black day today as this is the 2nd incident happening to my equipment, and Equallogic SAN was under very light load and the disk just failed without any pre-warning. Where is the predicative disk failure feature in Equallogic after all?

1

Alert from Dell OME:
Device: , Service Tag:, Asset Tag:, Date:12/12/12, Time:17:46:17:000, Severity:Critical, Message:Sent when eqlDiskStatus changes from one state to another state. Variables: eqlDiskStatus=Failed,eqlDiskSlot=5

Alert from EQL Group Manager:
Warning health conditions currently exist.
Correct these conditions before they affect array operation.
Non-fatal RAIDset failure. While the RAID set is degraded, performance and availability might be decreased. There are 1 outstanding health conditions. Correct these conditions before they affect array operation.
Failure: HDD Drive: 5, Model: ST3600057SS , Serial Number: 3SL14VVR
Reconstruction of RAID LUN 0 initiated.

Alert from SAN HQ

  • 12/12/2012 5:45:46 PM to 12/12/2012 5:47:46 PM
    • Warning: Member eql RAID Set Is Degraded
      • Warning: Member eql RAID set is degraded because a disk drive failed or was removed.
    • Warning: Member eql RAID More Spares Expected
      • Warning: Member eql The current RAID configuration requires more spare drives then are currently available.
    • Warning: Member eql has a failed drive in slot 5

Running Windows Server 2012 under ESX 4.1

By admin, December 4, 2012 13:10

I just saw this post, seemed someone got it working finally, I haven’t got time to try, if you did make it running, pls drop back a line, thanks.

Re: How to install windows 8 on esxi 4.1 ?

If you’d like to give it a shot, download the attached file and add the following options to your configuration file:

bios440.filename = “<full path to rom image>”
mce.enable = TRUE
cpuid.hypervisor.v0 = FALSE
vmGenCounter.enable = FALSE

Attachments:
bios.440.rom (512.0 K)

Solution:

Stop error 0×109: CRITICAL_STRUCTURE_CORRUPTION on a VMWare virtual machine

https://kb.vmware.com/s/article/2060019?sliceId=1&dialogID=158133794&docTypeID=DT_KB_1_1&stateId=0+0+158145165

To work around this issue, manually create a CPUID mask for the affected virtual machines:

To manually create a CPUID mask for the affected virtual machines:
  1. Power off the virtual machine.
  2. Right-click the virtual machine and click Edit Settings.
  3. Click the Options tab.
  4. Select CPUID Mask under Advanced.
  5. Click Advanced.
    • For Intel:
      • Under the Register column, locate the edx register under Level 80000001.
      • Paste this into the value field:—-:0—:—-:—-:—-:—-:—-:—-For example:

        cpuid.80000001.edx = —-:0—:—-:—-:—-:—-:—-:—-

Pages: Prev 1 2 3 4 5 6 7 8 9 10 ...24 25 26 Next