Chemical Databases

ChemSpider
Text or structure-based chemistry information, spectra and properties from the Royal Society of Chemistry.
NIST Chemistry WebBook
Thermochemical data and spectra from the National Institute of Standards and Technology.
USDA Nutrient Database
Energy content of a wide variety of foods from the United States Department of Agriculture. Enter keyword (e.g. peanut or corn puff) and press return. Then click on the report button (a 100 g sample size is convenient.)
ChemIDplus
Chemical structure, toxicity, and physical property searching from the National Library for Medicine.
FDA DailyMed
FDA labels and package inserts of marketed drugs for each brand name.
DrugBank
Drug data and drug target (i.e. sequence, structure, and pathway) information.
Chemical Entities of Biological Interest (ChEBI)
Dictionary of small chemical compounds from the European Bioinformatics Institute. Select the stars option (manually curated data.)
Pubchem Compounds
Free repository non-curated database of 30 million chemical molecules.

Other Sources

TotallySynthetic
Paul Docherty writes about a recent organic synthesis paper.
ChemSpider SyntheticPages
Interactive database of synthetic chemistry hosted at the Royal Society of Chemistry. "We publish practical and reliable organic, organometallic and inorganic chemical synthesis, reactions and procedures deposited by synthetic chemists. Synthetic methods on the site are updated continuously by chemists working in academic and industrial research laboratories. ChemSpider SyntheticPages encourages submissions from graduate students, postdocs, industrialists and academics."
Molbank
One compound per paper (Open Access Journal)
Org Prep Daily
The author goes by the name Milkshake. "The preparative procedures here don't have to be new - a good reproduction of original procedures or their slight modification is just as valuable. The criterion is reproducibility and general usefulness. The idea is that you did the procedure with your hands - preferably several times - and you have all experimental details available (NMRs, solvents used for crystallization, volumes, reaction times, etc.) and the procedure is useful to other chemists in medchem labs and academia. It is required that the procedure is robust enough that it will work in hands of an average chemist on the first try unless he does some blatant screw-up. (I hate simple-looking procedures with 81% published yield that later turns out to be 18% in reality)."

Journals