Tanking
This section will cover the subject of druid tanking as completely as I can: theory, practice, talents, and gear. This section is mostly my opinion. Obviously I think I’m right (otherwise it wouldn’t be my opinion) but remember that I am not the All-Knowing Bear God of Tanking.

Tanking Theory: What Do We Do?
A tank has two jobs in a group. The first job is to make sure that no mobs are hitting a healer or DPSer. When mobs are hitting people other than the tank it wastes time and resources at best and will lead to a wipe at worst. Tanks make sure that no mobs are hitting anybody other than tanks by generating the most threat of anybody in the fight (see below).

DIGRESSION: What is Threat and How Does It Work?
This section will explain what threat is and how it works. If you know this already, great. Otherwise, stick around. You just might learn something. Threat is what determines who a mob is attacking. A long time ago people used to think that mobs used some complicated AI to determine who they would attack, and there was no way we’d ever figure it out precisely. Turns out they were wrong on both counts. Mobs don’t use complicated AI, and we have figured it out pretty darn precisely.

Think of threat as an invisible resource you give to a mob by healing or buffing party members and doing damage to the mob. Whoever has given the mob the most threat is the one the mob attacks (see below for more details). If you hit the mob for 100 damage, you just generated 100 threat on that mob. If I hit the mob for 200 damage, I generated 200 threat on that mob. I have more threat than you, so it attacks me. If you hit it for 200 damage, you now have 300 threat on that mob, and it attacks you. Simple as that. Now let’s get more precise.

Every mob in the fight starts off with aggro on somebody. Aggro means who the mob’s target is; that’s why people say you “pulled aggro” if the mob stops attacking somebody else and starts attacking you. If nobody has generated any threat on a mob it will attack the nearest player (this is called “body aggro”). Otherwise it attacks the first person to generate threat against it. From that point forward the mob is on the threat mechanic.

Once a mob is on the threat mechanic (because at least one player has generated threat against it), there are two rules governing how to pull aggro. A player in melee range of the mob (i.e., the mob can melee that player without taking any steps) pulls aggro once it has 10% more threat than the player with aggro. So suppose the tank has 100 threat on the mob. Players in melee range can generate up to 109 threat and not pull aggro. At 110 threat, BAM, they’re the new target. And then the tank has to get up to 121 threat (10% more than 110) to get aggro back. This is known as the 10% rule, and applies to any player in melee range of a mob.

If the player is not in melee range, that player pulls aggro once it has 30% more threat than the player with aggro. So if the tank has 100 threat on the mob, a player at range could generate up to 129 threat without pulling aggro. It doesn’t matter how far away from the mob you are. 6 yards away or 100 yards away, the threshold is still 30%. This is known as the 30% rule and applies to any player not in melee range of a mob.

What about taunting? A taunt has up to three effects:

1.
It sets the taunter’s threat equal to whoever has aggro on the mob (note that this is not enough to trigger the 10% rule or the 30% rule.
2.
It gives the taunter aggro, violating the 10% rule and 30% rule.
3.
It gives the mob a short debuff that forces it to attack the taunter for as long as the debuff lasts, regardless of threat (i.e., it takes the mob temporarily off the threat mechanic).

Some abilities are taunt-like but not true taunts. For instance, the so-called “AOE taunts,” Challenging Roar and Challenging Shout, only have effect #3.

This is why the tooltip for Growl says it has no effect if the mob is already attacking you. Actually, strictly speaking, the tooltip is wrong. If you have 100 threat and aggro, and a hunter has 129 threat, and you Growl, you will still have aggro, but you will now have 129 threat instead of 100.

If the main way to generate threat on a mob is to hurt it or heal it, why don’t the DPSers and healers always have aggro? The answer is to cheat. All tanks have abilities that generate more threat than they otherwise would (and most DPSers and healers have abilities that generate less threat than they should, helping us out from the other end). If you try to keep aggro from a rogue by out-DPSing him, you’re going to fail. This is why it’s important for a tank to know which abilities generate the most threat for their rage cost (TPR): you’ve not only got to cheat, you have to cheat efficiently.

END DIGRESSION

Offensive Tanking Practice: How To Generate Threat
What separates a great tank from a tank that merely got you through the run? A great tank not only gets you to the end of the run, he gets you to the end of the run smoothly. A great tank finishes the instance in 45 minutes and leaves you going, “Wow … I want to run with that guy again!” How can a tank make a run go smoothly? Two ways.

1.
First, by being easy to heal. A great tank has geared in advance to make sure the healer isn’t always scrambling to keep him alive.
2.
Second, by having high, consistent threat per second (TPS). A great tank generates so much threat that the DPS can jump in right from the start, and the healers never even think about pulling heal aggro. A great tank knows how to use his rage to generate the maximum possible threat, resulting in not only high TPS but smooth TPS that isn’t threatened by a few big crits from a mage or a rogue.

So which abilities do generate the most threat? Turns out druids have it easy. There is one simple priority order for single-target tanking that all druid tanks must memorize for all mobs, bleedable or non-bleedable. In order from the most TPR to least: Mangle > Lacerate > Maul. In other words,

1.
Mangle if the cooldown is up.
2.
Lacerate if it isn’t.
3.
If you can get off one Mangle and three Lacerates every 6 seconds and still have rage left over, spam Maul until you’re out of rage while repeating 1-2.
4.
Repeat 1-3.

Swipe can replace Lacerate if you have extraordinarily high attack power (your Swipe has to land for about 336 after-armor damage before Swipe is better TPR than Lacerate).

Multi-mob tanking is somewhat harder. Swipe is our main tool for holding aggro on numeroase mobs. Swipe spam is good enough to hold aggro on three mobs in some circumstances. Swipe does not generate any extra threat (see Offensive Tanking Math, below). This means it generates threat purely through damage dealt. Depending on how hard you're getting hit, this might not be enough to out-aggro your healers. The usual way to do "Swipe spam" is to:

1.
Spam Swipe
2.
Spam Maul on your main target
3.
When tanking more than three mobs (which is difficult for a druid, but possible) change targets frequently to be sure all the mobs in the pull are getting Swiped.

Simple enough, if it works. If this is not enough to keep aggro on the off-targets, land some Lacerates (or even some Mangles) on the off-targets. A mouseover Lacerate macro can help, by casting Lacerate on your actual target unless your cursor is hovering over a live mob. This lets you use your Lacerate button like normal, but simply point at an off-target you want to Lacerate. Here's the basic macro:

#showtooltip
/cast [target=mouseover,nodead,harm][] Lacerate

Some offensive tanking tips and tricks:

1.
The first thing you hit your main target with after the initial pull should be a double Mangle/Maul combo unless you have a mega compelling reason not to. Mangle/Maul offers two benefits as an opening combo:
1.
It's the maximum amount of threat you can get in the space of a single hit. This gives you a nice threat lead over your DPS so they can open up immediately, which is worth violating the normal rule that says Maul is the last use of your rage.
2.
It gives you some insurance against misses, parries, and dodges - if one ability doesn't land, the second one almost certainly will, so you're virtually guaranteed to land with a high-threat ability right off the bat.
To perform a Mangle/Maul combo, target the mob and hit Maul before it's in range. This will cast Maul the instant the mob comes in range (which is also when you hit Mangle).

2.
Pick up 3/3 Intensity. You can then use Enrage before a pull and have enough rage to hit the mob immediately with a double Maul/Mangle combo. Your Enrage also then becomes a good tool for using mid-fight if you find yourself temporarily out of rage.

3.
If you do lose aggro, use Feral Charge to catch up with the mob. You can then Growl (which is a true taunt) to regain aggro immediately, and give the mob a good Maul/Mangle combo to make sure it doesn’t forget who it’s supposed to be hitting.

4.
Remember that Challenging Roar is not a true taunt. It forces the mobs off the threat mechanic, giving you temporary aggro for 6 seconds, but that’s it. If you don’t generate enough threat to get aggro on the mobs for real during those 6 seconds, they’ll go back to whatever they were doing when the 6 seconds are up. This makes Challenging Roar of limited use for trying to tank a huge pack of mobs.
*
On the other hand, if Growl is resisted, Challenging Roar can give you enough time to land a Maul/Mangle combo on the mob, which will probably boost you to the top of the threat list all by itself.
*
You can’t use Challenging Roar to get aggro on a huge pack of mobs, but you can use it effectively in AOE situations all the same. Wait until your AOEr is in trouble (e.g., the mage has blown all his snare cooldowns and the mobs are beating on him). Then hit Challenging Roar. Presto – the AOEr has 6 seconds of breathing room for a cooldown to finish, or just to get some more distance.
5.
There are a number of different ways to pull, all of which are valid. The different pulling strategies trade mana for extra initial threat, so you decide to balance maximal threat (and thus control of the fight) with drinking time. In order of increasing mana intensiveness:
*
From bear form, pop Enrage, and pull with Feral Faerie Fire.
*
Shift into bear form to get rage from Furor, pop Enrage, and pull with Feral Faerie Fire.
*
Starfire your main target, shift into bear form, pop Enrage, and hit with Feral Faerie Fire.
*
Wrath your off-target, Moonfire your main target while the Wrath bolt is still traveling, shift into bear form, pop Enrage, and hit with Feral Faerie Fire.
*
Hurricane an entire pack, shift into bear form, pop Enrage, and hit with Feral Faerie Fire.
6.
Demoralizing Roar does not generate a lot of threat. Do not use it as your primary tool to keep aggro against multiple mobs. However, it can be used to pick up a spawn or an add (something that is coming into the fight fresh, with no threat built up yet) without having to hit it.
7.
If you have a free special attack due to an Omen of Clarity proc, try to use it on Mangle or Lacerate rather than Maul. Remember that Maul costs about 8 rage in lost autoattack rage whether you pay 10 rage to use the ability or not, so OoC only saves you 10 rage on Maul, whereas it saves you 15 rage on Mangle and 13 rage on Lacerate. The difference is not critical, of course, and not worth seriously interrupting your TPS rotation for. However, if it just means delaying hitting the Maul key for a moment, it's worth it to save the extra rage.

Defensive Tanking Practice: How Not to Die
Now we know how to făurite high, consistent TPS. But how do we make ourselves easy to heal?

There are two things that will kill you as a tank: slow grinding damage and big spike damage. If you're dying to slow grinding damage it's because your long-term DPS intake from mobs is higher than your long-term HPS intake from your healers. This means that you don't have the gear to tank the fight, your healers don't have the gear to heal the fight, or both. Either way, stop trying until somebody gets some upgrades.

This means, in practice, a tank is going to die either to spike damage or to something that isn’t his fault (e.g., a healer fell asleep at the keyboard, so the tank got ground to death). There are three sources of big spike damage: crits, crushing blows, and spells. A crit deals 2X damage. A crushing blow deals 1.5X damage. A spell deals however much damage the spell deals, and can never crit or crush.

You should never die to crits (which deal 2x damage), because you can make yourself crit-immune (see mechanics section of part 1 of this guide for the exception). To make yourself crit-immune to a raid boss you need (as you know from the mechanics section) a total of -5.6% chance to be critted. So much for crits.

What about crushes? Warriors and paladins can become temporarily immune to crushing blows. Druids cannot*. You will always have a 15% chance to be crushed by a raid boss, no matter what you do. You cannot stack enough dodge, enough defense, or enough resilience to become crush immune. Your only defense against crushing blows is to stack armor and hit points enough that they don't matter. A crushing blow on a druid can and should be the same as a regular hit on a prot warrior. That’s your protection against crushes.

Your only defense against magic spike damage is AOE avoidance from Predatory Instincts (if the spell is AOE) or lots of hit points. Warriors and paladins innately take less damage than normal from spells; that’s their defense against magic spike damage. Ours is just to have lots and lots of HP.

Thus, the sources of survivability once you are crit-immune are:

1.
Dodge (from agility, defense, and dodge rating) - helps with slow grinding damage
2.
Armor (from armor) - helps with crushing blows and slow grinding damage**
3.
Hit Points (from stamina) - helps with all forms of damage

As a result of the above three bullet points, my [very strong] personal opinion is that a druid should stack armor and stamina first, and agility second. This doesn’t mean that dodge is bad. Dodge is good. Dodge is our friend. The more dodge you have, the less hard your healers have to work, and the more they will love you. Without enough dodge your slow grinding damage will be too much for your healers, and there’s nothing more discouraging than fighting your way to a boss only to discover you don’t have the gear for the fight.

HOWEVER, it also sucks to be three minutes into the boss fight and get two-shotted by back-to-back crushes because you didn’t stack enough armor and HP.

* Technically you could stack so much dodge that you would be crush-immune, if you had access to all the gear in the game. However, such a set would not be a viable tanking set. Its main problem is that you would dodge so often that you wouldn't have enough rage to keep aggro on the mob, so your DPS would be absurdly slow. If you had a really really well coordinated group you could probably do it, but in all likelihood your DPS would pull aggro just from autoattacks unless they deliberately downgraded their gear.

** I know what you’re thinking. You’re thinking that armor has diminishing returns, and dodge doesn’t. Please never say that out loud. Go read this instead.

Offensive Tanking Math: Threat
For those of you who want more detail, here are how the major druid abilities break down in terms of threat at level 70. If you want a more complete list for all the classes, see here:

*
Bear Form and Dire Bear Form multiply all base threat generated by 1.3. 5/5 Feral Instinct increases this multiplier to 1.45.
*
Mangle generates base threat equal to 1.3 times the damage inflicted. It is a druid’s highest-threat ability.
*
Lacerate does 285 base threat plus 20% of all damage it deals.
*
Maul does base threat equal to damage dealt plus 322. This makes it a druid’s second highest threat ability.
*
Swipe does base threat equal to damage dealt.
*
Demoralizing Roar generates base threat equal to 42.
*
Feral Faerie Fire and Faerie Fire generates base threat equal to 108.
*
All damaging spells generate base threat equal to the damage dealt.

Why is Maul third in priority even though it does more threat than Lacerate (and, seemingly, more than Mangle)? Even though Lacerate is the lowest threat of the three primary druid tanking abilities, it is second in priority because, being an instant attack instead of next melee, it is more threat per rage than Maul.

Let's assume that Mangle hits for 500.

*
Mangle hits for 1.15 times white damage plus 155, for 15 rage (talented). So if Mangle hits for 500, our white hits are for 300.
*
Maul adds 176 damage to the next attack, for 10 rage (talented). So our Maul would hit for 476.
*
A natural white hit from a bear of 300 damage generates 8.47 rage.

Now let's look at a 2.5 second window of time where we could either Mangle and get off one autoattack, or Maul once.

Mangle + Autoattack:

*
Mangle once for 500 x 1.3 = 650 base threat
*
Spend 15 rage
*
Autoattack once for 300 x 1.0 = 300 base threat
*
Gain 8.47 rage
*
Exhaustiv base threat generated: 950
*
Total rage spent: 6.53
*
Plenar base threat per rage: 145.48

Maul

*
Maul once for 476 + 322 = 798 threat
*
Spend 10 rage
*
Integral base threat generated: 798
*
Total rage spent: 10
*
Absolut base threat per rage: 89.8

For this reason the Mangle + autoattack combo will remain better total threat unless your autoattack hits for 200 or less. However, Mangle + autoattack is better TPR unless your autoattack hits for 140 or less (which is very low for a level 70 bear). So the rule of thumb for basically all bears is Mangle over Maul if you have to pick one. Similar math can be shown for Lacerate vs. Maul, which is why the rage-limited priority for a bear is Mangle > Lacerate > Maul.

Offensive Tanking Math: Rage
It's common to hear people talk about boss fights as giving a tank "unlimited rage." Obviously that's not true; all fights give a tank a finite and [theoretically] knowable amount of rage. But as you can now figure out, there's also a finite amount of rage that a druid can actually use. A "full on" single-target tanking rotation for a druid would be:

*
1 Mangle and 3 Lacerates every 6 seconds and
*
1 Maul every 2.5 seconds (assuming no haste effects)

This works out to exactly 14 rage per second (RPS). More rage than that is not bad per se, but it's not very useful either since at 14 rage per second a druid can be using a excepțional ability every single global cooldown and every single autoattack.

Knowing that 14 RPS is the magic "unlimited" number leads to some interesting observations. As you know, Primal Fury will grant 5 rage every time any bear attack crits, and all druid tanking attacks can crit (this is the most significant effect of Lacerate having an initial damage component - the crit might only hit for about 40 damage, but it still procs Primal Fury). If a bear has a crit chance of about 25%, this leads to an average of 1.33 rage per second (because a full-on bear lands an attack roughly once per second). That's 1.33 RPS from Primal Fury procs (more or less depending on crit chance).

What about rage from damage dealt? Maul generates no rage from damage, of course, but autoattacks generate a surprising amount. Here is the formula for rage generated through white damage dealt at level 70, which is quite simple:

[(7.5d / 274.4) + (fs)] / 2

where

*
d = damage dealt
*
f = 3.5 for a normal hit, 7.0 for a critical hit
*
s = weapon speed (always 2.5 for a bear assuming no haste)

So take a bear whose autoattack hits for an average of 300. That's

[(7.5 x 300) / 274.4 + (3.5 x 2.5)] / 2 = 8.5 rage

which in turn is 3.4 rage per second (actually 4.3 RPS when you factor in 25% crit chance). That's a significant amount given that the maximum RPS is only 14.0.

The formula for rage generated through damage taken at level 70 is even simpler:

2.5d / 274.7

Since you might be curious, this means that in order to gain 8.37 RPS (the amount our example druid needs to get up to the magic 14) our example druid would need to be taking 918 DPS exactly, after all defenses were taken into account.

Here's another interesting question: since a druid going all-out generates no rage from autoattacks (because every autoattack is a Maul), how much DPS would he need to take to reach 14, assuming only 1.33 RPS from Primal Fury procs? The answer (math omitted) is 1390 DPS, quite a bit more than if the druid isn't using Maul at all.

Oh, and in case you're feeling really hardcore, here's how to scale these formulas for lower levels than 70. Replace the 274.7 number with the following:

(0.0091107836 x Level^2) + (3.225598133 x Level) + 4.2652911

Tanking Talents
A full tanking build would look like this: http://www.wowhead.com/?talent=0ZxhGsfroeuioVxcz. There are some common feral talents missing from that build, but this is how I would maximize a druid for PVE tanking. The core tanking talents are:

1.
5/5 Ferocity – necessary to lower the cost of our abilities enough to keep up with realistic amounts of rage.
2.
3/3 Feral Instinct – changes the bear form threat multiplier from 1.30 to 1.45. That’s a huge boost.
3.
1/1 Feral Charge – this ability is our main way to catch up to mobs that got loose. Use it on mobs you lost aggro on, as an interrupt against spellcaster mobs, or against far away adds. A very versatile tool.
4.
2/2 Shredding Attacks – Lacerate is second in the single-target tanking rotation, so anything that reduces its rage cost is a must-have.
5.
2/2 Primal Fury – This is an important source of rage, and the key to our off-tanking prowess.
6.
1/1 Feral Faerie Fire – it doesn’t generate a lot of threat, but it does let you pull from bear form. And the armor reduction doesn’t hurt either.
7.
5/5 Heart of the Wild – +20% stamina is a lot. It also stacks multiplicatively with the bear form stamina modifier and Survival of the Fittest, meaning a fully talented bear gains 54% extra HP per point of stamina.
8.
3/3 Survival of the Fittest – the stat bonus is small, but the crit reduction is huge. You need this to become uncrittable.
9.
1/1 Mangle – our biggest, most efficient high-threat attack. You can tank without this but your TPS will drop dramatically, which is always dangerous.

Aldor or Scryer?
A word on the choice of the Aldor vs. the Scryers. Neither of the factions provides any gear that can't be replaced, and neither provides any recipes that you absolutely must have, unless your guild dictates it according to some Master Crafter Plan, and can't be persuaded after you copy and paste this section. The one thing they do provide that you can't replace or buy is inscriptions. This means in my opinion, for feral druids especially, that the choice of faction should predominantly be based on the choice the tanking inscriptions (the cat inscriptions are essentially interchangeable, but for what it's worth cats favor Scryers).

If you take a look at you'll see that the Scryer inscriptions favor defense and the Aldor inscriptions favor dodge:

*
Inscription of Warding (honored Aldor) - 13 dodge rating (0.69% dodge)
*
Inscription of the Knight (honored Scryer) - 13 defense rating (0.22% crit reduction, 0.22% dodge, 0.22% miss)
*
Greater Inscription of Warding (exalted Aldor) - 15 dodge rating (0.79% dodge), 10 defense rating (0.17% crit reduction, 0.17% dodge, 0.17% miss)
*
Greater Inscription of the Knight (exalted Scryer) - 15 defense rating (0.25% crit reduction, 0.25% dodge, 0.25% miss), 10 dodge rating (0.53% dodge)

All the inscriptions give exactly the same value in terms of item points. But as you can see, the Scryer inscriptions give significantly more crit reduction, and only the Scryer inscriptions give crit reduction at honored reputation.

It's true that the Aldor inscriptions give more long-term damage reduction. However, as you can see, if you can use the crit reduction, the Scryer inscriptions give nearly the same long-term DR. For this reason my personal preference for druid tanks is to go Scryer. The extra defense on the greater inscription can make the burden of finding crit reduction easier at the higher tiers of gear, and the fact that the regular inscription gives any defense at all means that a tank can be uncrittable with no defense gems even in all blues (i.e., even without Earthwarden). Of course, if you have a lot of resilience on your tanking gear and therefore don't need the extra defense from Scryer inscriptions, the Aldor inscriptions would have provided more damage reduction overall. You should consider where you plan to stop - if you don't ever expect to see your druid in full Tier 4, Aldor may be the way to go for you. But I consider the Scryer enchants the better choice for early tank gear (pre-Kara blues) and later tank gear (T4-T6).

As a side note, it's well worth grinding (or buying) yourself to exalted with your faction of choice, and especially worth it for the Aldor druid, to open up another source of defense. There are exactly six true druid tanking shoulders in the entire game (S1-3, T4-6), so it's not as if you'll be re-inscribing a constant stream of upgrades.

Tanking Gear
I don’t personally believe in tank points, or any other system that tries to rate tank gear on a single continuum of best to worst. The trouble with tank points is that they’re usually predicated on a long-term analysis so, e.g., 1% dodge means you take 1% less damage overall. That’s fine for a long-term analysis, but as I’ve already discussed, tanks get killed for short term reasons. However, if you must have a comprehensive listing of tank gear to use as a reference, go get Rawr.

As always, I advocate understanding the mechanics and making your own decisions. Here’s what the various tanking stats do for you:

1.
Dodge (from agility, defense, and dodge rating) - helps with slow grinding damage
2.
Armor (from armor) - helps with crushing blows and slow grinding damage
3.
Hit Points (from stamina) - helps with all forms of damage
4.
Attack Power (from strength and AP) – helps with TPS from autoattack and Mangle
5.
Crit chance (from agility and crit rating) – helps with TPS from all attacks and provides extra rage from Primal Fury procs
6.
Hit Rating (from hit rating) – helps with TPS by making all attacks less likely to miss, and makes Growl and Challenging Roar less likely to be resisted
7.
Expertise (from expertise rating) – helps with TPS by making all attacks less likely to be dodged and less likely to be parried. Twice as effective as hit rating at increasing TPS, but does not affect taunts.

A word on Hit Rating/Expertise.
While it is true that you do 0 TPS when dead, you might as well be dead if you can't keep aggro. Hit Rating and Expertise are relatively unimportant midway through the fight. DPS is usually conservative enough that it's trying to give you about a 10-20% TPS lead, which means that after a few minutes you've got enough of a raw threat lead to cover a string of misses, parries, and dodges (blocks will reduce the damage the mob takes, but the attack still "landed," meaning it still dealt damage and the innate threat of Lacerate and Maul was still applied).

Early in the fight is another matter. Early in the fight you don't have a raw threat lead to speak of (or else you're telling your DPS to hold off until you've established one, which is safe but tedious). The DPS will still be trying to give you a 10-20% TPS lead, but that's in the long term. In the short term you might get several misses/dodges/parries in a row, or they might get a few crits in a row, and then DPS has pulled aggro in the short term, and you aren't doing your job.

Is it your fault? Well, kind of. You can't control their crits, but you can control your misses through Hit Rating and Expertise. How much do you need? Depends on your tolerance for risk, but consider the following:

*
The 10% and 30% rules mean aggro is easier to keep than to regain. The easiest way to keep aggro is never to lose it.
*
With 0 HR and 0 Expertise the odds of Maul or Mangle (the usual big-threat openers) not landing against a raid boss (=miss/dodge/parry) is roughly 20%. But the odds of both missing is about 4%. This is why a Maul/Mangle combo opener is so important.
*
After the initial opener you probably don't have the rage for constant Maul/Mangle or Maul/Lacerate combos (if you do, then great; you're pretty well insulated). That means, without Hit Rating or Expertise, you're dealing with a 20% chance that you'll go for a few seconds without a single yellow attack landing.

The cheapest way to reduce that ugly 20% number is to stack some Expertise. But Hit Rating is a lot easier to find. I myself limitează the 20% number to about 15% (same as pelf, if you read the comments below) and find that feels a lot more controllable than tanking "raw."

A Word About Gems
It is an Inviolate Druid Rule not to gem for defense until you have at least enchanted your bracers with defense and your chest with resilience. My strong personal preference is to also inscribe your shoulders for defense before using gem sockets for defense.

You can't really consider the question of gems without also considering the question of enchants at the same time. Consider the following alternatives:

1.
Head -Glyph of Ferocity vs. Glyph of the Defender
2.
Chest - Major Resilience vs. Heavy Knothide Armor Kit or Exceptional Stats
3.
Bracer - Major Defense vs. Fortitude or Stats
4.
Sockets - +8 defense vs. +8 agility or +12 stamina or +4 agility/+4 hit rating

Notice a couple things about the list up there:

*
Your chest enchant slot gets you either 0.38% crit reduction or 10 stamina
*
Your bracer enchant slot gets you either 0.20% crit reduction or 12 stamina
*
Your sockets get you either 0.13% crit reduction or 12 stamina

Stats in sockets are pretty much interchangeable; none of the tanking gems let you beat the system (e.g., 2x +4agi/+6stam = 1x +8 agi + 1x +12 stam, and 8 defense rating costs the same as 8 agility or 12 stamina). However, not all enchants are created equal. From a tanking perspective, +12 defense to bracers gets you a lot more than +stats to bracers, as does +resilience to chest compared to a heavy knothide armor kit. Which would you rather have? +8 defense from a gem and +10 stamina from your chest? Or +15 resilience from your chest and +12 stamina from a gem? The answer is obviously the latter.

I prefer the tanking inscriptions to the DPS ones. Neither crit rating nor attack power are critical stats for tanks, whereas the defense from inscriptions can free up gem sockets to carry stats that are more critical, like agility, hit rating, or stamina.

Glyphs are another question, since Glyph of Ferocity offers hit rating (which is good, but hard to come by) in exchange for two sockets' worth of defense. Personally I decided on my glyph last, and managed to come up with enough defense without wasting gems (or even enchant slots) that I could take Glyph of Ferocity (since I'd rather have that hit rating than some extra dodge).

As for which gems to pick, my personal opinion is to go for Solid Star of Elune all the way if you can, but you already know I have a personal bias for stamina over agility. I also have a personal bias for armor over both, so my personal tanking gear tends to run low on stamina but high on armor, which I try to make up for with extra stamina gems. Remember though that unless you must gem for defense, your sockets should only be spent on:

*
Stamina
*
Agility
*
Hit Rating

The întocmai combination of gems is, as I said, immaterial. Unlike healing gems, which let you "beat the system" if you go with combination colors, all possible combinations of tanking gems work are equal in terms of item point values and sockets required.