Batch, 256 bytes

@set i=echo 
@%i%off
set u=
for /l %%j in (2,2,%1)do call set i=%%i%%  
set/af=%1^&1
if %f%==1 %i% __&%i%/,,\o&set u=__
for /l %%j in (2,2,%1)do call:l
exit/b
:l
set i=%i:~0,-2%
%i%  _%u%_
%i%o/,%u%,\
%i% __%u%__
%i%/,_%u%_,\o
set u=__%u%__

Note that line 1 has a trailing space and line 4 has two trailing spaces. i therefore contains an echo command with the appropriate amount of indentation for each turtle. Meanwhile u contains the number of underscores in alternate turtles. A leading odd turtle is special-cased and then the rest of the turtles are output in pairs.

Cheddar, 105 bytes

n->(|>n).map(i->(1-i%2)*"o"+"\\,"+(n-i-1)*"__"+",/"+i%2*"o"+i/2*"  "+"\n"+(n-i)*"__"+(i+1)*" ").vfuse.rev

Ruby, 100 bytes

Recursive solution. Try it online!

f=->n,i=1{f[n-1,i+1]if n>1;puts' '*i+?_*n*2,"%#{i-1}s/,#{?_*2*~-n},\\"%(i<2?'':'o '[i%2])+' o'[i%2]}

V, 57 53 bytes

bdwi __
/,,\o@"ñHyjç,/f,2á_k2A_
ëPhjI ñLòkk$x^PXkk

Since this contains unprintable characters, here is a hexdump:

00000000: 1862 6477 6920 5f5f 0a2f 2c2c 5c6f 1b40  .bdwi __./,,\o.@
00000010: 22f1 4879 6ae7 2c2f 662c 32e1 5f6b 3241  ".Hyj.,/f,2._k2A
00000020: 5f0a eb50 1668 6a49 20f1 4cf2 6b6b 2478  _..P.hjI .L.kk$x
00000030: 5e50 586b 6b                             ^PXkk

Try it online!

This should be 56 bytes, since there is H is supposed to be mapped to gg, which worked in previous versions, but magically disappeared from this version. :/

Explanation:

<C-x>                       "Decrement the input
     bdw                    "Move back and delete the input into register '"'
        i __\n/,,\o<esc>    "Insert the first turtle

@"ñ                         "Input - 1 times:
   H                        "  Move to the beginning of the file
    yj                      "  Copy two lines
      ç,/                   "  On every line containing a comma: 
         f,                 "    Move the cursor to that comma
           2á_              "    Append two underscores after the comma
              k             "    Move up a line
               2A_          "    And append 2 underscores after this line

ë               "  Move to the beginning of the file
 P              "  Paste the two lines we previously copied
  <C_v>hjI      "  And indent these two lines with a space
           ñ    "End loop

L               "Move to the end of the file
 ò              "  Recursively:
  kk            "  Move up two lines
    $x          "  Delete the last character on this line (an 'o')
      ^P        "  And paste this 'o' at the beginning of the line
        X       "  Remove one space
         kk     "  Move up two lines again

C, 328 bytes:

main(m,e,O,y,p,F,g,z,s){scanf("%d",&m);e=m;O=2*e;for(y=2;y<=2*m;y+=2){p=O%4==0;char x[y-2],d[e],l[e-1-p],u[y];x[y-2]=d[e]=l[e-1-p]=u[y]='\0';for(F=0;F<e-1-p;++F)l[F]=' ';for(g=0;g<y;++g)u[g]='_';for(z=0;z<(y-2);++z)x[z]='_';for(s=0;s<e;++s)d[s]=' ';printf("%s%s\n%s%s/,%s,\\%s\n",d,u,(O>4)? l:"",p ? "o":"",x,p ? "":"o");e--;O-=2;}}

A simple implementation using a lot of string formatting and for loops. However, another, hopefully shorter, recursive solution is coming soon. I will still try and golf this current one as much as I can, though.

C It Online! (Ideone)

Python 2, 153 bytes

n=input()
s=' ';r=[];k=0;i=0
while i<n:a=i%2;b=1-a;r=[s*k+s*a+s+'_'*(n-i)*2+s,s*k+'o'*a+'/,'+'_'*(n*2-i*2-2)+',\\'+'o'*b]+r;k+=a*2;i+=1
print'\n'.join(r)

Try it online

Retina, 97 bytes

Byte count assumes ISO 8859-1 encoding.

.+
$*_
_(_*)
 /,$1$1,\o
+`^( *)(.)(/.+)__,.(.)
 $1$4$3,\$2¶$&
%`^( *)./,(_*)
$1  $2__¶$&
Rm`^ | $

Try it online!